The American Journal of Psychiatry in 1980 (2)

Volume 162, Issue 7

In This Issue:

In This Issue Am J Psychiatry 2005 162: A48. Perspectives

Editorials:

Robert M.A. Hirschfeld Are Depression and Bipolar Disorder the Same Illness? Am J Psychiatry 2005 162: 1241-1242.

Reviews and Overviews:

Kenneth S. Kendler "A Gene for...": The Nature of Gene Action in Psychiatric Disorders Am J Psychiatry 2005 162: 1243-1252.

Images in Neuroscience:

Michael Vogel The Cerebellum Am J Psychiatry 2005 162: 1253.

Introspections:

Jeanne Steiner In Sickness and Health Am J Psychiatry 2005 162: 1254. Images in Psychiatry:

Gregg E. Gorton Milton Hyland Erickson, 1901–1980 Am J Psychiatry 2005 162: 1255. New Research

Articles:

Jean A. Frazier, Sufen Chiu, Janis L. Breeze, Nikos Makris, Nicholas Lange, David N. Kennedy, Martha R. Herbert, Eileen K. Bent, Vamsi K. Koneru, Megan E. Dieterich, Steven M. Hodge, Scott L. Rauch, P. Ellen Grant, Bruce M. Cohen, Larry J. Seidman, Verne S. Caviness, and Joseph Biederman Structural Brain Magnetic Resonance Imaging of Limbic and Thalamic Volumes in Pediatric Bipolar Disorder Am J Psychiatry 2005 162: 1256-1265. Maria E. Fisfalen, Thomas G. Schulze, J. Raymond DePaulo, Jr., Leslie J. DeGroot, Judith A. Badner, and Francis J. McMahon Familial Variation in Episode Frequency in Bipolar Affective Disorder Am J Psychiatry 2005 162: 1266-1272. Ralph W. Kupka, David A. Luckenbaugh, Robert M. Post, Trisha Suppes, Lori L. Altshuler, Paul E. Keck, Jr., Mark A. Frye, Kirk D. Denicoff, Heinz Grunze, Gabriele S. Leverich, Susan L. McElroy, Jörg Walden, and Willem A. Nolen Comparison of Rapid-Cycling and Non-Rapid-Cycling Bipolar Disorder Based on Prospective Mood Ratings in 539 Outpatients Am J Psychiatry 2005 162: 1273-1280. Mauricio Tohen, Waldemar Greil, Joseph R. Calabrese, Gary S. Sachs, Lakshmi N. Yatham, Bruno Müller Oerlinghausen, Athanasios Koukopoulos, Giovanni B. Cassano, Heinz Grunze, Rasmus W. Licht, Liliana Dell’Osso, Angela R. Evans, Richard Risser, Robert W. Baker, Heidi Crane, Martin R. Dossenbach, and Charles L. Bowden Olanzapine Versus Lithium in the Maintenance Treatment of Bipolar Disorder: A 12-Month, Randomized, Double-Blind, Controlled Clinical Trial Am J Psychiatry 2005 162: 1281-1290. Ramin Mojtabai, Daniel Herman, Ezra S. Susser, Nancy Sohler, Thomas J. Craig, Janet Lavelle, and Evelyn J. Bromet Service Use and Outcomes of First-Admission Patients With Psychotic Disorders in the Suffolk County Mental Health Project Am J Psychiatry 2005 162: 1291-1298. Glenn N. Saxe, Frederick Stoddard, Erin Hall, Neharika Chawla, Carlos Lopez, Robert Sheridan, Daniel King, Lynda King, and Rachel Yehuda Pathways to PTSD, Part I: Children With Burns Am J Psychiatry 2005 162: 1299-1304.

Julie B. Kaplow, Kenneth A. Dodge, Lisa Amaya-Jackson, and Glenn N. Saxe Pathways to PTSD, Part II: Sexually Abused Children Am J Psychiatry 2005 162: 1305-1310. Miles McFall, Andrew J. Saxon, Charles E. Thompson, Dan Yoshimoto, Carol Malte, Kristy Straits-Troster, Evan Kanter, Xiao-Hua Andrew Zhou, Cynthia M. Dougherty, and Bonnie Steele Improving the Rates of Quitting Smoking for Veterans With Posttraumatic Stress Disorder Am J Psychiatry 2005 162: 1311-1319. Axel Perkonigg, Hildegard Pfister, Murray B. Stein, Michael Höfler, Roselind Lieb, Andreas Maercker, and Hans-Ulrich Wittchen Longitudinal Course of Posttraumatic Stress Disorder and Posttraumatic Stress Disorder Symptoms in a Community Sample of Adolescents and Young Adults Am J Psychiatry 2005 162: 1320-1327. Mark Olfson, Marc J. Gameroff, Steven C. Marcus, Ted Greenberg, and David Shaffer National Trends in Hospitalization of Youth With Intentional Self-Inflicted Injuries Am J Psychiatry 2005 162: 1328-1335. Ann F. Garland, Anna S. Lau, May Yeh, Kristen M. McCabe, Richard L. Hough, and John A. Landsverk Racial and Ethnic Differences in Utilization of Mental Health Services Among High-Risk Youths Am J Psychiatry 2005 162: 1336-1343. Susan S.F. Gau, M.Y. Chong, Tony H.H. Chen, and Andrew T.A. Cheng A 3-Year Panel Study of Mental Disorders Among Adolescents in Taiwan Am J Psychiatry 2005 162: 1344-1350. Joseph R. Calabrese, Paul E. Keck, Jr., Wayne Macfadden, Margaret Minkwitz, Terence A. Ketter, Richard H. Weisler, Andrew J. Cutler, Robin McCoy, Ellis Wilson, Jamie Mullen, and The BOLDER Study Group A Randomized, Double-Blind, Placebo-Controlled Trial of Quetiapine in the Treatment of Bipolar I or II Depression Am J Psychiatry 2005 162: 1351-1360. Research Units on Pediatric Psychopharmacology Autism Network Risperidone Treatment of Autistic Disorder: Longer-Term Benefits and Blinded Discontinuation After 6 Months Am J Psychiatry 2005 162: 1361-1369.

Brief Reports:

Mark Zimmerman, Iwona Chelminski, and Michael A. Posternak Generalizability of Antidepressant Efficacy Trials: Differences Between Depressed Psychiatric Outpatients Who Would or Would Not Qualify for an Efficacy Trial Am J Psychiatry 2005 162: 1370-1372. Patrik K.E. Magnusson, David Gunnell, Per Tynelius, George Davey Smith, and Finn Rasmussen Strong Inverse Association Between Height and Suicide in a Large Cohort of Swedish Men: Evidence of Early Life Origins of Suicidal Behavior? Am J Psychiatry 2005 162: 1373-1375. Alexandre Dumais, Alain D. Lesage, Aleksandra Lalovic, Monique Séguin, Michel Tousignant, Nadia Chawky, and Gustavo Turecki Is Violent Method of Suicide a Behavioral Marker of Lifetime Aggression? Am J Psychiatry 2005 162: 1375-1378. Justine M. Kent, Jeremy D. Coplan, Osama Mawlawi, Jose M. Martinez, Susan T. Browne, Mark Slifstein, Diana Martinez, Anissa Abi-Dargham, Marc Laruelle, and Jack M. Gorman Prediction of Panic Response to a Respiratory Stimulant by Reduced Orbitofrontal Cerebral Blood Flow in Panic Disorder Am J Psychiatry 2005 162: 1379-1381. Richard Meiser-Stedman, William Yule, Patrick Smith, Ed Glucksman, and Tim Dalgleish Acute Stress Disorder and Posttraumatic Stress Disorder in Children and Adolescents Involved in Assaults or Motor Vehicle Accidents Am J Psychiatry 2005 162: 1381-1383. Communications and Updates

Letters to the Editor:

SERMSAK LOLAK, ELISA DANNEMILLER, and FRANCIS ANDRES 48,XXYY Syndrome, Mood Disorder, and Aggression Am J Psychiatry 2005 162: 1384. DANIEL T. HOLMES, PHILLIP LONG, and JIRI FROHLICH Dysbetalipoproteinemia and Clomipramine Am J Psychiatry 2005 162: 1384-1385. ROBERT OSTROFF, MARBIELA GONZALES, and GERARD SANACORA Antidepressant Effect of Ketamine During ECT Am J Psychiatry 2005 162: 1385-1386. TUBA OZCANLI, BARIS UNSALVER, SAMURAY OZDEMIR, and MINE OZMEN Sertraline- and Mirtazapine-Induced Severe Neutropenia Am J Psychiatry 2005 162: 1386. RAKESH KARMACHARYA, MARI MINO, and WILLIAM F. PIRL Clozapine-Induced Eosinophilic Colitis Am J Psychiatry 2005 162: 1386-1387. JEAN ADDINGTON and DONALD ADDINGTON Clinical Trials During the Prodromal Stage of Schizophrenia Am J Psychiatry 2005 162: 1387. COLIN O’DONNELL and TIM STEPHENS The Significance of Homocysteine Levels in Schizophrenia Am J Psychiatry 2005 162: 1387-1388. DONALD C. GOFF Dr. Goff and Colleagues Reply Am J Psychiatry 2005 162: 1388. ANDREW J. DWORK, GORAZD ROSOKLIJA, and LIESL B. JONES Reduced Spinophilin in Schizophrenia Am J Psychiatry 2005 162: 1389. AMANDA J. LAW, CYNTHIA SHANNON WEICKERT, THOMAS M. HYDE, JOEL E. KLEINMAN, and PAUL J. HARRISON Dr. Law and Colleagues Reply Am J Psychiatry 2005 162: 1389-1390. MARK S. GOLD, KIMBERLY FROST-PINEDA, and RICHARD J. MELKER Physician Suicide and Drug Abuse Am J Psychiatry 2005 162: 1390. EVA SCHERNHAMMER Dr. Schernhammer Replies Am J Psychiatry 2005 162: 1390. RYAN M. CARNAHAN and PAUL J. PERRY Methodological Concerns in a Trial of Ziprasidone and Olanzapine Am J Psychiatry 2005 162: 1391. DAVID E. ROSS Methodological Concerns in a Trial of Ziprasidone and Olanzapine Am J Psychiatry 2005 162: 1391. GEORGE M. SIMPSON Dr. Simpson Replies Am J Psychiatry 2005 162: 1391-1392. LIEUWE de HAAN and NICO van BEVEREN Olanzapine and Haloperidol for Residual Symptoms Am J Psychiatry 2005 162: 1392-1393. ROBERT W. BUCHANAN, M. PATRICIA BALL, and ELAINE WEINER Dr. Buchanan and Colleagues Reply Am J Psychiatry 2005 162: 1393. JACK C. SCHOENHOLTZ Origin of the Term "Schizophrenia" Am J Psychiatry 2005 162: 1393.

Book Forum:

SAXBY PRIDMORE The Neurological Basis of Pain Am J Psychiatry 2005 162: 1394. FRANCINE M. BENES Neurology for the Non-Neurologist, 5th ed. Am J Psychiatry 2005 162: 1394-1395. ROBERT HOWARD Neuropsychiatric Assessment Am J Psychiatry 2005 162: 1395. RICHARD B. MAKOVER Sleep Medicine in Clinical Practice Am J Psychiatry 2005 162: 1395-1396. DAVID V. FORREST Brain Dynamics and Mental Disorders: Project for a Scientific Psychiatry Am J Psychiatry 2005 162: 1396-1397. MARK H. FLEISHER The Mind: Its Nature and Origin Am J Psychiatry 2005 162: 1397. MARY V. SEEMAN The Philosophy of Psychiatry: A Companion Am J Psychiatry 2005 162: 1397. WILLIAM M. GREENBERG The Biopsychosocial Approach: Past, Present, Future Am J Psychiatry 2005 162: 1398. LAURA L. POST Mind, Meaning, and Mental Disorder: The Nature of Causal Explanation in Psychology and Psychiatry, 2nd ed. Am J Psychiatry 2005 162: 1398-1399. TRUCE T. ORDOÑA Feeling Good: The Science of Well-Being Am J Psychiatry 2005 162: 1399. DONALD M. HILTY, BLYTHE A. CORBETT, and PIERRE LAVENEX Affect Dysregulation and Disorders of the Self Am J Psychiatry 2005 162: 1399-1400. GORDON PARKER Experiences of Depression: Theoretical, Clinical, and Research Perspectives Am J Psychiatry 2005 162: 1400.

THE AMERICAN JOURNAL OF In This Issue PSYCHIATRY

Brains of Bipolar Kids Kids Hospitalized for Self-Injury

Compared to adults with bipolar disorder, Between 1990 and 2000, the overall rate children and adolescents with this illness of hospitalization for children and adoles- have more severe symptoms, a stronger cents who intentionally injured them- genetic predisposition, and less past drug selves declined only slightly. However, the treatment. Their brains might therefore average length of these hospital stays de- provide a clearer picture of anatomic ab- creased significantly, and diagnoses shifted normalities in bipolar disorder. Frazier et toward more severe conditions, e.g., bipo- al. (p. 1256) measured three brain regions lar disorder. Olfson et al. (p. 1328) derived in 43 youths with bipolar disorder and 20 these findings from over 10,000 hospital- healthy youths. The amygdala and thala- izations for self-inflicted injuries of pa- mus showed no difference between groups, tients ages 5–20 years. Although the rate but the hippocampus, which provides con- did not change overall, hospitalizations in- text for learning and memory, was smaller creased for children ages 5 to 9 years, for in the bipolar youths. The bipolar girls ingestion of acetaminophen and antide- showed an especially pronounced differ- pressants, and for injuries from cutting ence from the same-sex comparison sub- and from hanging or suffocating. Cutting jects, and their mania scores were higher may be more closely related to self-mutila- than those of the bipolar boys. The total tion than to suicidal behavior, however, cerebral volume in the bipolar group was and without it the rate of hospitalizations also smaller than normal. Since total cere- for self-inflicted injuries shows an actual bral volume generally is normal in affected Milton Hyland Erickson, 1901–1980 decline, from 47.2 per 100,000 population adults, bipolar disorder in children and ad- Images in Psychiatry (p. 1255) in 1990 to 39.4 per 100,000 in 2000. olescents may follow a different neurodevelopmental path. Race of High-Risk Youths Affects Use of Mental Health Services The Incredible Shrinking Hospital Stay Many children and adolescents in child welfare programs, the From 1989 to 1995, the average length of hospitalization for juvenile justice system, and other public care sectors need men- psychotic disorders in Suffolk County, N.Y., plummeted from 41 tal health services. The Surgeon General’s Report on Mental to 21 days. Nevertheless, Mojtabai et al. (p. 1291) found that Health has identified racial/ethnic disparities in use of mental posthospitalization symptoms and illness course and outcomes health services as a major public health problem. Garland et al. were no worse among patients admitted in 1995 than among (p. 1336) examined whether ethnicity itself, apart from associ- those hospitalized in 1989. There was also no significant change ated factors such as socioeconomic status, influenced the use of in the number of rehospitalizations, the rate of suicide or mental health services by 1,256 children and teens in five types homelessness, or use of outpatient mental health services. In of public care in San Diego County. Compared to non-Hispanic fact, patients admitted in the later years were more likely to whites, African Americans and Asian Americans were approxi- have returned to their highest level of functioning by 4-year fol- mately half as likely to receive outpatient services, as well as low-up. Introduction of atypical antipsychotic drugs did not ac- mental health services in general. Other predictors, such as count for this lack of decline in patient outcomes. One troubling caregiver strain, were also identified and suggest that attention finding was the greater likelihood of partial or no remission at to family- and system-level factors could open doors to young discharge among patients hospitalized later in the study period. people needing help. Linkage to outpatient services is especially important for these patients.

80 Non-Hispanic white (N=554) Protecting Burned Children From PTSD Latino American (N=332) 70 African American (N=282) Fire is a common cause of accidental injuries to children. Little Asian American/ 60 is known about which of these children develop posttraumatic Pacific Islander (N=88) stress disorder (PTSD) and which do not. Saxe et al. (p. 1299) 50 found two separate causal pathways to PTSD in 72 children admitted to a hospital for burn victims. Separation anxiety and 40 dissociation of mental processes (e.g., amnesia for traumatic events) independently contributed to PTSD at 3 months after 30 hospitalization. Each was itself influenced by the size of the burn. Separation anxiety was also greater for younger children Percent Who Used Service 20 and those with greater pain. Anxiety and dissociation appear to reflect different biological reactions to overwhelming stress—a 10 fight-or-flight response versus freezing/immobilization. These two pathways to PTSD suggest specific interventions to help 0 burned children: increased availability of their parents in the Outpatient 24-hour care Informal Any mental hospital, support of the parents, and psychosocial and biologi- services health service cal treatments for pain, anxiety, and dissociation. Service Category

A48 Editorial

Are Depression and Bipolar Disorder the Same Illness?

A long-standing scientific controversy with many clinical consequences is whether bipolar and unipolar disorder are the same or separate and distinct illnesses. Two articles in this issue of the Journal present findings that address this controversy with new evidence. Depression has long been viewed as a fundamental human condition. Descriptions of the syndrome of depression are included in the writings of Hippocrates from 2,500 years ago. He attributed melancholic temperament to an excess of black bile emanating from the liver, one of four bodily humors (1). The symptoms of dysphoria, psychomotor retardation, and suicidality are consistent with DSM-IV criteria. In contrast, the concept of bipolarity as a fundamental human condition is quite new. Although “mania” was mentioned by ancient Greek physicians, including Hippocrates, its description varied widely, with little consistency (2). In the late 19th and early 20th century, Kraepelin, in Germany, emphasized the distinction between dementia praecox and manic-depressive insanity (3, 4). One source of this dichotomy was the emphasis on moods versus cognition and will. However, the major rationale for the distinction was a perceived differ- “These findings support ence in clinical course and outcome. Dementia the separation of bipolar praecox was characterized by a deteriorating course, whereas mood disorder insanity was disorder from recurrent characterized by a relapsing course (5). His con- unipolar illness.” cept of manic-depressive insanity included both recurrent unipolar and bipolar illness. It was not until 1966 that bipolar disorder was described as separate and distinct in articles by Jules Angst (6), Carlo Perris (7), and Winokur and colleagues (8) in the United States. These articles proposed separate unipolar and bipolar disorders based on difference in genetics, gender, clinical course, and premorbid personality. The two articles in the current issue of the Journal add to the growing literature supporting the distinctness of bipolar disorder. In the article by Fisfalen and colleagues, the authors report that the frequency of illness episodes is highly familial among patients with bipolar disorder compared with recurrent unipolar depression. This finding serves to provide evidence of a genetic difference between bipolar disorder and unipolar disorder. The article by Frazier and colleagues also addresses the issue of bipolar disorder as a distinct illness with a biological basis. The authors found structural differences in components of the limbic system in prepubertal children with bipolar disorder. Since regulation of mood and basic human drives (e.g., sleep, sex, and appetite) is a function of the limbic system, these structural differences relate to causes of the illness. A very intriguing finding in the study by Frazier et al. is that of decreased hippocampal volumes in children with bipolar disorder, which has not been reported in adults. This raises the question of whether prepubertal bipolar disorder is somehow different from adolescent- or adult-onset bipolar disorder. Consistent with this is a difference in the presenting clinical practice of mania in children (9). Instead of a euphoric mood, the most frequent mood disturbance is severe irritability, often with accompanying protracted, hostile, and violent temper outbursts (10). During these affective storms, it is nearly impossible to calm the child. This situation tends to be persistent rather than ep- isodic (as is seen more often in adults). At the same time, classic symptoms of mania oc-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1241 EDITORIAL

cur in these children. These include grandiosity, hypersexuality, increased energy, and decreased need for sleep. It is intriguing to speculate whether these hippocampal imaging differences between children on one hand and adolescents and adults on the other are responsible for the differences in clinical presentation.

References

1. Papadimitriou GN, Dikeos DG, Soldatos CR: The concept of bipolar disorder: a historical perspective, in Handbook of Bipolar Disorder: Diagnosis and Therapeutic Approaches. Edited by Kasper S, Hirschfeld RMA. New York, Marcel Dekker (in press) 2. Angst J, Marneros A: Bipolarity from ancient to modern times: conception, birth and rebirth. J Affect Disord 2001; 67:3–19 3. Kraepelin E: Psychiatrie, 4. Leipzig, Germany, JA Barth, 1893 4. Kraepelin E: Psychiatrie, 5. Leipzig, Germany, JA Barth, 1896 5. Angst J: Historical aspects of the dichotomy between manic-depressive disorders and schizophrenia. Schizophr Res 2002; 57:5–13 6. Angst J: Zur Aetiologie und Noslogie endogener depressiver Psychosen. Berlin, Springer, 1966 7. Perris C: A study of bipolar (manic-depressive) and unipolar recurrent depressive psychoses. Acta Psychiatr Scand 1966; 42(suppl 194):1–189 8. Winokur G, Clayton PJ, Reich T: Manic Depressive Illness. St Louis, CV Mosby, 1969 9. Biederman J, Mick E, Faraone SV, Spencer T, Wilens TE, Wozniak J: Pediatric mania: a developmental subtype of bipolar disorder? Biol Psychiatry 2000; 48:458–466 10. Wozniak J, Biederman J, Richards J: Diagnostic and therapeutic dilemmas in the management of pediatric- onset bipolar disorder. J Clin Psychiatry 2001; 62(suppl 14):10–15

ROBERT M.A. HIRSCHFELD, M.D.

Address correspondence and reprint requests to Dr. Hirschfeld, Department of Psychiatry and Behavioral Sciences, University of Texas Medical Branch, 301 University Blvd., Galveston, TX 77555; [email protected] (e- mail).

1242 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Reviews and Overviews

“A Gene for…”: The Nature of Gene Action in Psychiatric Disorders

Kenneth S. Kendler, M.D. A central phrase in the new “GeneTalk” is of these criteria. The concept of “a gene “X is a gene for Y,” in which X is a particu- for…” is best understood as deriving from lar gene on the human genome and Y is a preformationist developmental theory in complex human disorder or trait. This ar- which genes—like preformationist anla- ticle begins by sketching the historical or- gen—“code for” traits in a simple, direct, igins of this phrase and the concept of the and powerful way. However, the genetic gene-phenotype relationship that under- contribution to psychiatric disorders fails lies it. Five criteria are then proposed to to meet any of the five criteria for the con- evaluate the appropriateness of the “X is cept of “X is a gene for Y.” The impact of a gene for Y” concept: 1) strength of individual genes on risk for psychiatric ill- association, 2) specificity of relationship, ness is small, often nonspecific, and em- 3) noncontingency of effect, 4) causal bedded in complex causal pathways. The proximity of X to Y, and 5) the degree to phrase “a gene for…” and the preforma- which X is the appropriate level of expla- tionist concept of gene action that under- nation for Y. Evidence from psychiatric ge- lies it are inappropriate for psychiatric netics is then reviewed that address each disorders.

(Am J Psychiatry 2005; 162:1243–1252)

The last 20 years has seen the rise of “GeneTalk” (1). A phenotype that it causes—originated in the developmen- central phrase in GeneTalk, and one that has been heard tal theory of preformationism (5). One of the earliest artic- widely in both lay (2) and professional arenas, is “X is a gene ulated theories of development, preformationism was first for Y,” in which X is a particular gene on the human genome proposed by Aristotle but became particularly influential and Y is one of a wide variety of complex human disorders in the 17th century (3, 5, 6). The essentials of the theory are or traits such as depression, aggression, sexual orientation, eloquently described by Jacob: obesity, infidelity, alcoholism, or schizophrenia. This essay begins with a brief review of the historical ori- At a time when living beings are known by their vis- gins of the concept of “a gene for…”. I then propose criteria ible structure alone, what has to be explained about to assess the validity of this model of gene-phenotype rela- generation [i.e., development] is the maintenance of this primary structure through succeeding generations and go on to evaluate these criteria as applied to ge- tions. The structure cannot itself disappear; it has to netic effects on psychiatric disorders. The essay concludes persist in the seed from one generation to another. To with general observations about our preconceptions and maintain the continuity of shape, the “germ” of the lit- the reality of gene action in psychiatric disorders. Although tle being to come has to be contained in the seed; it many of the issues raised in this essay are equally applica- has to be “preformed.” The germ already represents the visible structure of the future child.…It is the plan ble to etiologically complex medical disorders, the focus for the future living body…already materialized, like a here will be on psychiatric illness. miniature of the organism to come. It is like a scale model with all the parts, pieces and details already in Historical Origins of the Concept position.…Fertilization only activates it and starts it growing. Only then can the germ develop, expand in of “A Gene for…” all directions and acquire its final size, like those Japa- nese paper flowers which, when placed in water, un- Since humans started speculating about the nature of wind, unfold and assume their final shape. (7, p. 57) development and inheritance, a number of different conceptualizations have emerged about the nature of the In preformationism, the egg or sperm was understood guiding forces in these processes (3). In the 20th century, to contain all the final traits of the mature organism. De- this discourse has come to focus largely on the nature of velopment consisted of the expansion of these preformed what Mendel originally termed “anlagen” or “elements,” characteristics (or anlagen) into the individual traits of the which in 1909 became “genes” (4). adult organism. That is, these anlagen were truly for the Of the multiple different views of the nature of the “gene,” adult traits with which they had a simple and direct causal the one in which we are interested—a gene defined by the relationship.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1243 GENE ACTION IN PSYCHIATRIC DISORDERS

In the 19th century, as the young field of biology strug- While medical geneticists came to understand that in gled to fathom the mechanism of transmission of traits biological systems, genes actually code for proteins, it be- across generations, a number of the proposed theories of came convenient and seemingly natural to think about inheritance (where the “units” of inheritance had names preformationist-like genes for these classical genetic dis- such as pangenes, stirps, and gemmules) had important eases in humans. preformationist themes (3, 4). When Mendel’s ground- The last 30 years have seen three interrelated further breaking work on genetics (originally published in 1866) themes in the “a gene for…” story. First, in the mid-1970s was rediscovered in 1900, one common interpretation was two influential books appeared that heightened the profile that his “elements of inheritance” were the discrete an- of genes and their potential impact on human behavior. lagen predicted by preformationist theories (5). This inter- “Sociobiology: The New Synthesis” by Wilson (13) pretation was favored by two of the most influential genet- launched the field of sociobiology (and later evolutionary icists of the day, the Dutchman de Vries (the most famous psychology), discourse in which commonly included the of the three “co-rediscovers” of Mendel [4]) and the En- concept of “genes for” a wide range of traits, including al- glishman Bateson (8). truism, territoriality, jealousy, and ethics. “The Selfish In summarizing this exciting period in the history of bi- Gene” by Dawkins (14) proposed a gene-centered view of ology, Allen (9) writes evolution in which an organism, with its wide array of phenotypes, was viewed as a vehicle through which genes replicate themselves over evolutionary time. Second, with The implications that the discreteness of the gene implied the organism was constructed as a “mosaic” the developmental of an ever increasing set of powerful of adult traits was given explicit voice by Bateson with molecular tools, the specific genes and then the specific the first years of his encounter with Mendelism. mutations in those genes were discovered that were responsible for all major classic human genetic disorders. Allen goes on to quote two passages from Bateson writ- So, when speaking about “a gene for Y” in which Y was ten, respectively, in 1901 and 1902 (9): sickle-cell anemia, cystic fibrosis, or Huntington’s chorea it became possible to conceive of the gene not only as an In so far as Mendel’s law applies, the conclusion is abstract transmitted “unit” but also as a discrete piece forced upon us that the living organism is a complex of DNA at a specific location on a chromosome. Third, of characteristics of which some, at least, are dissocia- prompted by the sequencing of the human genome, the ble and are capable of being replicated by others. We thus reach the conception of unit characters which concept that DNA represented the “blueprint” of life (or in may be rearranged in the formation of reproductive related versions the “code” or “recipe” for life) was widely cells. promulgated in both the scientific and lay literature (2). The organism is a collection of traits. We can pull The preformationist themes in this metaphor are evident: out the yellowness and plug in greenness, pull out genes are to phenotypes as blueprints of a building are to tallness and plug in dwarfness. the building themselves. So, this historical sketch suggests that our current con- Bateson was recasting, in a new language, preforma- cept of “X is a gene for Y” in humans has four major inter- tionist concepts. The Mendelian anlagen (later genes) related historical roots. First, the concept that develop- could be defined by their relationship to the particular ment anlagen could be “for” adult traits arose in phenotype (or “unit character”) with which it had a privi- preformationist developmental theories. Second, the dis- leged causal link. That is, such genes caused phenotypes in covery of Mendel’s “elements” was interpreted by some as the same way that the preformationist anlagen prefigured verifying this concept. Third, the idea that genes could be adult traits. From this perspective, it made sense to speak “for” human traits was supported by the discovery that of “a gene for greenness,” “a gene for tallness,” or a gene for genes for classical Mendelian medical disorders often any of the other innumerable unit characteristics of the acted just like the hereditary elements found in Mendel’s adult organism. It is in this context that a rarely discussed pea plants. Finally, these concepts became linked to DNA early chapter of psychiatric genetics in the United States by a series of stunning discoveries in the last 20 years, so must be viewed, when reports appeared claiming to find, that strength of the “icon” of the double helix provided in series of large pedigrees, evidence for Mendelian genes particular luster to potential discoveries in psychiatry of “a “for” “Nomadism or the wandering impulse” (10) and “the gene for…”. neuropathic constitution” (11). This preformationist concept of the gene proved attrac- Criteria for the Concept tive to medical geneticists who, over the course of the 20th of “A Gene for…” century, showed that most classical genetic disorders in humans (termed “Mendelian” diseases in honor of the The remainder of this essay addresses the question of Austrian monk) were due to hereditary units that behaved whether this preformationist model of gene action—in just like those first examined by Mendel (12). which genes are “for” phenotypes—is appropriate for psy-

1244 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KENNETH S. KENDLER chiatry. Based in part on prior efforts to develop guidelines There is no way to “acquire” cystic fibrosis or Huntington’s for causal inference in epidemiology (e.g., reference 15), I disease through environmental exposure. So if having the suggest five criteria by which to judge the validity of the disease gene always produces the disorder and the disor- claim “X is a gene for Y”: 1) strength of association of X with der never occurs without the disease gene, this produces a Y, 2) specificity of relationship of X with Y, 3) noncontin- perfect association between the disease gene (X) and the gency of the effect of X on Y, 4) causal proximity of X to Y, disorder (Y). (Reality is somewhat more complex. Most and 5) the degree to which X is the appropriate level of ex- Mendelian genes in man contain several different muta- planation for Y. In sum, I argue that tions, each of which can cause diseases that are sometimes of quite variable severity. But this claim still holds for If gene X has a strong, specific association with dis- all mutations of the gene considered together.) ease Y in all known environments and the physiologi- The strength of an association between a risk factor and cal pathway from X to Y is short or well-understood, a disease is most frequently quantified by a statistic called then it may be appropriate to speak of X as a gene for Y. the odds ratio. Formally, the odds ratio is defined as the ra- But first, a few details are needed. The scientific basis of tio of the odds of developing the disease among those ex- most claims that “X is a gene for Y” results from a statisti- posed to the risk factor and the odds of disease among cal test called association analysis. In its simplest form, those not exposed to the risk factor. For Mendelian disor- this test compares the frequency of specific DNA variants ders in man, since the first of these figures is one and the in or around gene X in a set of cases with disorder Y and a second is zero, the odds ratio for the disorder given the set of matched control subjects. An association is claimed pathogenic gene is infinite. Since this is a rather stringent if the frequency of these variants differ significantly in criteria, for the sake of argument, let us say the association cases and control subjects. In both a conceptual and sta- with Mendelian-like genes (an historical model for the tistical sense, this approach is no different from the meth- concept of “a gene for…”) has an odds ratio of approxi- ods commonly used in the biomedical and social sciences mately 100 (Figure 1). to assess the relationship between putative risk factors Are there any genes whose strength of association with a and outcome variables such as smoking and lung cancer psychiatric disorder is Mendelian-like? Two related sources or childhood sexual abuse and depression. of information, both gathered in the last two decades, indi- Therefore, standard “a gene for…” claims are based on cate that the answer to this question is almost certainly statistical and not biological grounds. Biological studies “No.” First, a gene that has a deterministic or nearly deter- that trace etiologic pathways from X to Y should follow ministic relationship with a phenotype produces an un- claims for association and would certainly provide confir- mistakable signature in the pattern of illness in large pedi- matory data. However, they have been very rare to date in grees. Numerous investigators have now searched many psychiatric genetics. On its own, a significant p value in an parts of the globe (including nearly all psychiatric facilities association study tells you nothing about the nature of the in a modest-sized country [16]) seeking pedigrees in which causal relationship between the gene and the disease. major forms of psychiatric illness—especially schizophrenia and bipolar illness—are distributed in the pattern ex- Strength of Association pected from a Mendelian-like gene. Such pedigrees have As with any risk factor for any outcome, the strength of not been found. association between a specific gene and a particular dis- Second, Mendelian-like genes also produce a distinctive ease can vary in magnitude. In considering the criteria for result in genome-wide linkage studies, which effectively “a gene for…,” an historical standard of comparison is sweep the human genome looking for regions that contain what has come to be called a Mendelian gene. The action genes that have an impact on risk of illness. While the of Mendelian genes is deterministic and not probabilistic. technical details need not concern us, experts agree that If a plant inherits a particular copy of the gene for wrinkled for those disorders studied in genome-wide linkage scans peas, it would not matter how much sunshine the plant re- of reasonable size and quality—especially schizophrenia, ceived or the quality of its fertilizer. The plant will have bipolar illness, panic disorder, and eating disorders—con- wrinkled peas no matter what the environment does. In clusive evidence has accumulated that even moderately humans, we have many diseases that are due to Mendelian rare genes of Mendelian-like effect do not exist. (The avail- genes that behave exactly like the genes Mendel studied in able evidence does not permit us to rule out, however, very his pea plants (12). If you have one copy of the pathogenic rare Mendelian-like genes.) gene for Huntington’s disease, it does not matter what So, if we lack Mendelian genes for psychiatric disorders, your diet is, whether your parents were loving or harsh, or with their very high odds ratios, what sort of magnitude of if your peer group in adolescence were boy scouts or petty associations might we expect? One set of benchmarks criminals. If you have the mutated gene and you live long might be provided by three examples of what would be con- enough, you will develop the disease. sidered very strong associations in epidemiology. The esti- Furthermore, for most Mendelian genes in man, the mated odds ratio between heavy smoking and small cell only way to get the disorder is to have the disease gene. carcinoma of the lung is approximately 20 (17), between in-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1245 GENE ACTION IN PSYCHIATRIC DISORDERS

FIGURE 1. A Comparison of Estimated Odds Ratios for the TABLE 1. Meta-Analysis Results Published Since 2000 for Strength of Association Between Risk Factors and Key Studies of Association Between Individual Genes and Outcomesa Psychiatric Disorders

100 Odds Disorder Gene/Markera Ratio Schizophrenia (22) DRD3 S9G (SS) 1.13 90 5-HT2AR 102 T/C 1.07 Bulimia (23) 5-HT2AR 1438 G/A 1.33 Anorexia nervosa (23) 5-HT2AR 1438 G/A 1.42 Schizophrenia (25) DRD2 Ser311Cys 1.43 80 Bipolar illness (26) 5-HTT 44 base pair insertion 1.14 5-HTT intron 2 variable number of tandem repeats 70 polymorphism 1.18 Attention deficit DRD5 dinucleotide repeat hyperactivity disorder (27) polymorphism 1.57 60 DAT1 40 base pair variable number of tandem repeats polymorphism 1.27 DRD4 48 base pair variable 50 number of tandem repeats polymorphism 1.41 Odds Ratio a DR=dopamine receptor; 5-HTT=serotonin transporter; DAT= 40 dopamine transporter; 5-HT2AR=serotonin 2A receptor.

30 ciation for psychiatric disorders: Mendelian-like (odds ratio of approximately 100), strong (odds ratio=12–20), or moderate (odds ratio=3–10). 20 Trying to summarize the magnitude of association found between functional candidate genes and psychiatric dis- 10 orders is problematic because of the multiple method- ologic difficulties in the interpretation of such studies (22– 0 24). Greatest reliability should be placed on the results of Perfect Strong Moderate Modest meta-analyses, which are now beginning to appear in the (Mendelian-like (psychiatric genes) disorders) literature. A PubMed search from 2000 on (using publica- Strength of Association tion type of “metaanalysis” and search words “gene” and a Although the odds ratio for a classic Mendelian gene is actually ∞, “association”) followed by a hand search and elimination we estimate it here at approximately 100. Strong association (here of duplication yielded 10 significant meta-analytic esti- odds ratio=15) approximates that seen between heavy smoking and lung cancer, industrial exposure to asbestos and mesothe- mates of odds ratios between individual genes and psychi- lioma, and severe stressful life events and the onset of major de- atric disorders (Table 1) (excluding results from those pression. Moderate association (odds ratio=5.0) approximates that meta-analyses that did not support the original positive seen for apolipoprotein E-4 and Alzheimer’s disease as well as the protective effect in Asian populations of the ALDH2*2 copy of the reports). The odds ratios ranged from 1.07 to 1.57 with a aldehyde dehydrogenase gene on risk for alcoholism. The associa- median of approximately 1.30. tions seen between individual genes (or high-risk haplotypes) and Another strategy to localize candidate genes is to look psychiatric disorders (odds ratio=1.5) is an approximation obtained from a review of the current literature. for them under linkage peaks (so-called positional candidate genes). In schizophrenia, replicated evidence is now emerging for several such genes (28). For these genes, dis- dustrial exposure to asbestos and mesothelioma is approx- ease-associated haplotypes—small sections of DNA that imately 15 (18), and between severe stressful life events and have traveled together over evolutionary time—can often the onset of major depression is approximately 12 (19). be found. The two best replicated positional candidate Another more modest benchmark is provided by the two genes for schizophrenia are dysbindin 1 and neuregulin 1. outstanding genetic association results in neuropsychiatry Not counting the original reports (where the effect size of the last decades. The association between the pathogenic might be biased upward), estimates are available for the “4 allele” of the apolipoprotein E gene and Alzheimer’s dis- association between high-risk haplotypes and schizo- ease produces, in Caucasian populations, an odds ratio of phrenia for both of these genes. For dysbindin, odds ratios approximately 3.0 (20). In Asian populations, the posses- of 1.24 (29), 1.23 (30), 1.40 (31), 1.70 (32), and 1.58 (33) sion of the slow-metabolizing (ALDH2*2) copy of the alde- have been reported or calculated from replication reports. hyde dehydrogenase gene conveys up to a 10-fold reduc- For neuregulin 1, two replications were noted in a recent tion in risk for the development of alcoholism (21). review, with odds ratios estimated to be 1.25 and 1.80 (28). So, as depicted in Figure 1, we have three possible Taken together, the meta-analyses of functional candi- benchmarks for the strength of the gene-phenotype asso- date gene association studies and early results from posi-

1246 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KENNETH S. KENDLER tional candidate genes suggest that the magnitude of the FIGURE 2. Possible Gene-to-Phenotype Relationshipsa associations between individual genes and psychiatric ill- One-to-one One-to-many nesses have small odds ratios, largely from 1.1 to 1.6. Com- relationship relationship pared to our benchmarks, this effect size is very modest (Figure 1). Perhaps genes (or particular mutations or hap- V lotypes) of larger effect size will be found. While results W from linkage studies suggest that this is unlikely, it cannot XYX be ruled out. Also to be considered is the statistical dictum Y that the first set of effects detected in any research area tend to be the most robust. If this is correct, further genes Z discovered for psychiatric disorders are likely to have smaller average effects than the genes found to date. Many-to-one Many-to-many relationship relationship The preformationist concept of “a gene for…” implied a predetermined and largely irrevocable link between gene A and phenotype. This is the pattern of association observed A V between gene and phenotype from Mendel’s original traits B B W and for Mendelian genetic disorders in humans. By con- Y trast, for psychiatric disorders, individual genes appear to X X Y have a quite modest association with psychiatric illness. C C Z While they may have an impact on risk, individual genes hardly predetermine illness, as would be expected if we a Possible relationships between genes on the left-hand side of the had discovered “genes for” mental disorders. figure and phenotypes on the right-hand side are shown. In a one- to-one relationship, gene X causes only phenotype Y, and pheno- Specificity of Association type Y is caused only by gene X. In a one-to-many relationship, gene X causes several phenotypes each in turn being only caused by X. In The second criterion to evaluate the appropriateness of a many-to-one relationship, phenotype Y is caused by several genes each in turn only causing Y. In a many-to-many relationship, each the concept of “X is a gene for Y” is the degree of specificity gene causes several phenotypes and each phenotype is caused by in the relationship between X and Y. As illustrated in Fig- several genes. ure 2, does X influence risk for any other disorders in addition to Y? Or are there other genes that contribute to Y in disorders have been demonstrated in other twin studies addition to X? (e.g., references 35–37). In preformationist theory, anlagen had highly specific We know much less about the specificity of the spec- associations with the adult traits into which they devel- trum of effects on psychiatric disorders of individual oped. The hereditary elements of the pea that Mendel genes. Meta-analyses reviewed in Table 1 show that vari- studied also had quite specific phenotypic effects. That is, ants at one gene (the 5-HT receptor) may predispose to one gene influenced pea color but not shape or height 2A risk for three different disorders (schizophrenia, bulimia, while another influenced shape but not height or color. and anorexia nervosa). A pair of overlapping genes on However, as genetics developed, many genes were found chromosome 13q (termed G30 and G72) may be associ- that impacted on a variety of phenotypic characteristics— ated both with schizophrenia and bipolar illness (28). A a phenomenon called pleiotropy. number of overlapping positive regions in linkage genome In man, many Mendelian genes produce one and only scans for bipolar illness and schizophrenia have led some one disease syndrome (although sometimes of varying se- to argue that this reflects shared genes between these two verity depending on the specific mutation). But there are exceptions where different abnormalities in a single gene disorders (38). While difficult to evaluate critically, claims can produce distinct genetic diseases. have been made that several popular candidate genes (e.g., serotonin transporter, dopamine transporter, dopa- How specific are individual genes in their impact on risk mine 2 receptor) are significantly associated with a wide for psychiatric disorders? Do most genes influence risk for one and only one psychiatric disorder? Twin studies, variety of psychiatric disorders or psychiatrically relevant which study “genes” in the aggregate, suggest that genetic traits (39, 40). While much remains unknown, current evi- risk factors for psychiatric disorders are often nonspecific dence suggests that many genes that influence risk for in their effect. A large-scale twin study of seven psychiatric psychiatric disorders will not be diagnostically specific in and substance use disorders found one common genetic their effect, thereby resembling the one-to-many relation- risk factor predisposing to drug abuse, alcohol depen- ship in Figure 2 rather than the one-to-one relationship. dence, antisocial personality disorder, and conduct disor- We are on firmer ground in evaluating whether genetic der and a second common genetic factor influencing risk risk for psychiatric disorders results from the action of a for major depression, generalized anxiety disorder, and single gene (the one-to-one relationship in Figure 2) or phobia (34). Overlap of genetic risk factors for multiple multiple genes (the many-to-one relationship in Figure 2).

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1247 GENE ACTION IN PSYCHIATRIC DISORDERS

Some evidence bears on this question indirectly, as fol- the etiology of posttraumatic stress disorder. In the afore- lows. Twin and adoption studies provide convincing evi- mentioned multivariate twin model, what distinguished dence for significant genetic effects on virtually all major major depression, generalized anxiety disorder, and pho- psychiatric disorders (41). Therefore, genes that affect risk bia from one another were environmental and not genetic for these disorders must exist somewhere on the human risk factors (34). In a detailed study of the impact of child- genome. Linkage studies examine how these aggregate ge- hood parental loss on risk for common psychiatric and netic risk factors are distributed across the genome. If ge- substance use disorders, death of a parent was specific in netic risk resulted from a single gene, then all the linkage increasing risk for major depression and no other disorder “signal” would be concentrated in a single location, with a (Kendler et al., unpublished results). Consistent with stud- resulting clear and robust statistical linkage peak. But, as ies of stressful life events that have shown moderate sepa- noted earlier, this is a pattern that has not been observed ration of depressogenic and anxiogenic events (45, 46), a in published genome scans for psychiatric disorders. In- multivariate genetic study of symptoms of anxiety and de- stead, a number of modest linkage peaks are usually seen, pression showed that genetic factors influence nonspe- suggesting that the “packets” of genetic risk for these dis- cific risk for all symptoms, whereas two environmental orders are widely dispersed across the genome. (To com- factors were identified that predisposed, with moderate plicate matters, genome scans will underestimate the specificity, for symptoms of depression and anxiety, re- number of genomic regions involved because of low power spectively (47). to detect genes of small effect size, but will overestimate The preformationist concept of “a gene for…” implies the number because some of the observed “peaks” will be high levels of specificity between gene and phenotype. false positives.) While much remains to be learned in this area, current ev- Recently, data have emerged that addresses this ques- idence suggests that instead of the “one-to-one” relation- tion directly. A careful meta-analysis of 20 genome scans ship implied by the concept of “a gene for…,” genes and for schizophrenia has suggested 10 genomic regions likely disorders in psychiatry are likely to have the “many-to- to contain susceptibility genes (42). In addition, current many” relationship depicted in Figure 2. evidence of bipolar disorder, the second-best-studied psy- (The evidence that the association between individual chiatric disorder by linkage scans, also suggests multiple genes and psychiatric disorders are typically weak and loci (43). may often be nonspecific does not mean that the identifi- The specificity of association implied in the “a gene cation of such genes is unimportant. For example, such for…” concept has another implication worth exploring. discoveries can identify pathophysiologic pathways, begin Consistent with preformationist theory, specificity of gene the lengthy process of clarifying how individual genes in- action implies that the gene contains all information teract with each other and with environmental exposures needed for the development of the trait. The environment to produce illness, and provide new targets for treatment.) might impact on the final phenotype, but its effect is nonspecific. That is, the gene “codes for” the trait, while the en- Noncontingency of Association vironment reflects background factors that support devel- Noncontingent association means that the relationship opment but is not in and of itself “information-carrying.” between gene X and disorder Y is not dependent on other To illustrate how commonly we see genes and environ- factors, particularly exposure to a specific environment or ment in this light, it is worth pondering a curious and on the presence of other genes. As mentioned earlier, this asymmetrical feature of GeneTalk. While we find it easy to is a typical (albeit not uniform) feature of genes that cause use the phrase “X is a gene for Y,” it feels quite odd to say “A classical Mendelian disorders in humans. If the associa- is an environment for B.” For example, a large body of em- tion between gene and disease were contingent on partic- pirical work supports the hypothesis that severe life events ular environmental exposures, then we would have to are important environmental risk factors for major de- amend our statement to read “X is a gene for Y given expo- pression (44). The magnitude of the association between sure to environment Z.” such events and the subsequent depressive episode is far Environmental contingencies for genetic effects on psy- greater than that observed for any of the genes that we chiatric disorders have been little investigated. Twin and have reviewed here. Yet, who has heard the phrase “a ro- adoption studies suggest that the impact of aggregate mantic breakup is an environment for depression”? I sug- “genes” for major depression are altered by exposure to gest that we feel comfortable with “X is a gene for Y” and stressful life events (19, 48) and for schizophrenia and con- not “A is an environment for B” because we implicitly as- duct disorder by exposure to a dysfunctional rearing envi- sume that genes have a privileged causal relationship with ronment (49, 50). A range of twin studies suggest that envi- the phenotype not shared by environmental factors. ronmental experiences have an impact on genetic risk for However, empirical evidence does not support the posi- several psychiatrically relevant traits, including aggres- tion that genes code specifically for psychiatric illness sion, disinhibition, and smoking (51). Recently, Caspi and while the environment reflects nonspecific “background colleagues have found evidence for interactions between effects.” By definition, environmental factors are central to environmental risk factors and particular genes in the pro-

1248 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KENNETH S. KENDLER duction of antisocial behavior (52) and depression (53), Another vignette: with the former finding having been replicated (54). Assume a Mendelian genetic disease due to a muta- We know almost nothing about gene-by-gene interaction in gene K. Gene K’s normal function is to produce an tions in the etiology of psychiatric disorders. Although a enzyme L that breaks down metabolite M in cells allow- number of association studies have reported interactions, ing M to be harmlessly secreted from the body. When K I am unaware of any that have been widely replicated or has a pathogenic mutation, the enzyme L that is pro- supported by meta-analyses. Using statistical models ap- duced no longer works. Therefore, levels of M rise, pro- plied to risk of illness in various classes of relatives, Risch ducing a well understood series of toxic effects, thereby producing the genetic disorder N. has claimed that gene-by-gene interactions are important in the etiology of schizophrenia (55). This scenario suggests the following potentially simple Overall, we know little about the contingent nature of causal chain: mutated gene K→dysfunctional enzyme genetic effects for psychiatric disorders. The available in- L→excess metabolite M→disorder N. In this admittedly formation suggests that gene action contingent upon cer- oversimplified story, a case could be made that gene K had tain environmental exposures is probably not rare and sufficient causal proximity to disorder N to make plausible may be relatively common for psychiatric disorders. This the claim that “K is a gene for N.” However, it might be ar- is also inconsistent with the preformationist concept of “a gued that even here, the complexity of the paths from lev- gene for…”. els of M to disorder N may be far from “simple.” Contrast this situation to the causal chain from a gene Causal Proximity mutation to a complex psychiatric disorder such as Preformationist developmental models assumed that schizophrenia. Although early efforts have been made to anlagen developed directly into adult traits. The “blue- begin to trace such pathways (e.g., reference 56), we prob- print for life” metaphor similarly assumes a direct corre- ably do not know enough to articulate all the specific spondence between individual parts of the blueprint causal steps that would be needed to go from DNA base- (windows, doors, fixtures) and the corresponding units of pair variation to, for example, the cognitive processes that the completed building. Conceptualizing genes in this predispose to delusion formation. What we can conclude preformationist framework therefore carries the implicit with some confidence is that it will be very complex. In- assumption of a direct causal link between gene and phe- deed, the causal link between that hydraulic cable and the notype. It is only with this assumption that usage of the “a jumbo jet flying will probably look very simple and short gene for…” is congruent with the common meaning of the compared to the causal relationship between individual phrase “X is for Y” in English. To clarify this point, let’s ex- genes and the manifestations of schizophrenia. While the amine a typical list of such statements: nature of the evidence reviewed here is largely inferential, it suggests that the pathways from most genes for psychi- I use a knife for buttering my toast. atric illness to their phenotypes would fail the causal prox- I have a backpack for carrying my computer to work each day. imity criterion implicit in the concept of “X is a gene for Y.” I was upset at my son for not doing his chores. Appropriate Level of Explanation In each case, there is an implied direct and immediate Scientific theories typically strive to explain phenome- relationship between X and Y. To put it more formally, X non at the most informative level. To provide an absurd and Y are directly linked in a formal logical train of action example, no one would seek to understand the origin of (first two examples) or thought (third example). hypertension at the level of quarks. In some ultimate way, Now, how does this common sense meaning of the word quarks may be involved. But quarks are just the wrong “for” apply to the phrase “X is a gene for Y”? Let me illus- level of inquiry for the problem. trate the problem with a vignette To illustrate how this issue—the appropriateness of level of explanation—may apply to our evaluation of the concept A jumbo jet contains about as many parts as there are genes in the human genome. If someone went into the of “a gene for…” consider these two “thought experiments”: fuselage and removed a 2-foot length of hydraulic cable Defects in gene X produce such profound mental re- connecting the cockpit to the wing flaps, the plane could tardation that affected individuals never develop not take off. Is this piece of equipment then a cable for speech. Is X is a gene for language? flying? A research group has localized a gene that controls de- Most of us would be uneasy answering yes to this ques- velopment of perfect pitch (57). Assuming that individu- tion. Why? Because this example violates our conception of als with perfect pitch tend to particularly appreciate the causal proximity. When we say X is for Y, we expect X to be, music of Mozart, should they declare that they have found a gene for liking Mozart? to a first approximation, directly and immediately related to Y. That is not the case for the cable and flying. There are For the first scenario, the answer to the query is clearly many, many mechanical steps required to get from the “No.” Although gene X is associated with an absence of function of that cable to a jumbo jet rising off the runway. language development, its phenotypic effects are best un-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1249 GENE ACTION IN PSYCHIATRIC DISORDERS derstood at the level of mental retardation, with muteness psychiatric disorders. I then reviewed the available evi- as a nonspecific consequence. X might be a “gene for” dence, which was of variable quality, that addressed each mental retardation but not language. of these criteria. Although the second scenario is subtler, if the causal The strength of association between individual genes pathway is truly gene variant→pitch perception→liking and psychiatric disorders is weak and often nonspecific. Mozart, then it is better science to conclude that this is a Genes do not appear to contain all the information needed gene that influences pitch perception, one of the many ef- for the development of psychiatric illness, since environ- fects of which might be to alter the pleasure of listening to mental factors have, for several disorders, been shown to Mozart. It is better science because it is more parsimoni- have causal specificity. The action of genes on psychiatric ous (this gene is likely to have other effects such as influ- disorders may frequently be contingent on environmental encing the pleasure of listening to Haydn, Beethoven, and exposures, although much needs to be learned in this area. Brahms) and because it has greater explanatory power. The causal chain from genes to psychiatric disorders is A final scenario: probably long and complex. The appropriate level of explanation for gene action is much more likely to be basic bio- Scientist A studied the behavioral correlates of a par- logical or mental processes that contribute to psychiatric ticular variant at gene X and concluded “This is a very interesting gene that increases the rates of sky diving, disorders rather than the disorders themselves. Thus, with speeding, mountain climbing, bungee jumping, and un- varying degrees of confidence, the genetic contribution to protected casual sex.” Scientist B studied the same vari- psychiatric disorders fails to meet any of the five criteria for ant and concluded “This is a very interesting gene and the preformationist concept of “a gene for…”. The impact effects levels of sensation-seeking.” of individual genes on risk for psychiatric illness is small, Who has done the better science? Since sensation seeking often nonspecific, and embedded in causal pathways of (and its close cousin novelty-seeking) are well studied traits stunning complexity. (41), scientist B has provided results that are more parsimo- On this basis, I suggest that we conclude that the phrase nious and potentially provide greater explanatory power. “X is a gene for Y,” and the preformationist concept of gene For example, only scientist B could predict that this gene action that underlies it, are inappropriate for psychiatric ought to be related to other behaviors, like drug taking, that disorders. The strong, clear, and direct causal relationship are known to be correlated with sensation-seeking. implied by the concept of “a gene for…” does not exist for As reviewed here, genes have been and will continue to psychiatric disorders. Although we may wish it to be true, be found that have statistical relationships with risk for we do not have and are not likely to ever discover “genes psychiatric disorders. However, will the action of these for” psychiatric illness. genes be best explained at the level of the disorders themselves? While we cannot answer this question definitively, Received June 18, 2004; revision received Aug. 24, 2004; accepted Sept. 21, 2004. From the Virginia Institute for Psychiatry and Behav- I would judge this to be unlikely. Far more plausible is that ioral Genetics, Departments of Psychiatry and Human Genetics, Med- we will find genes whose mode of action can be best un- ical College of Virginia of Virginia Commonwealth University, Rich- derstood at the level of more basic biological processes mond. Address correspondence and reprint requests to Dr. Kendler, Department of Psychiatry, P.O. Box 980126, Richmond, VA 23298- (e.g., neuronal cell migrations during development) and/ 0126; [email protected] (e-mail). or mental functions (e.g., processing of threat stimuli). Supported by a Fritz Redlich Fellowship from the Center for Ad- vanced Study in the Behavioral Sciences and the Rachel Brown Banks Endowment Fund. Overview and Conclusion The author thanks Kenneth Schaffner, M.D., Ph.D., Ralph Green- span, Ph.D., Brien Riley, Ph.D., and Douglas Levinson, M.D., for their The goal of this essay is to understand the historical ori- review of an earlier version of this essay. gins of the key phrase “X is a gene for Y” and then to evaluate its appropriateness for psychiatric disorders. Our in- References terest, of course, is not merely the phrase itself, but the conceptual framework that underlies this form of Gene- 1. Kitcher P: The Lives to Come: The Genetic Revolution and Hu- Talk. The use of the phrase “a gene for” implies (and in fact man Possibilities. New York, Simon & Schuster, 1996 only makes sense in the context of) genes which—like pre- 2. Nelkin D, Lindee MS: The DNA Mystique: The Gene as a Cultural Icon. New York, WH Freeman, 1995 formationist anlagen—“code for” psychiatric illness in a 3. Magner L: A History of the Life Sciences, 2nd ed. New York, simple, direct, and powerful way. Marcel Dekker, 1994 I argue that the concept of “a gene for…” can best be un- 4. Dunn L: A Short History of Genetics. New York, McGraw-Hill, derstood as deriving from preformationist developmental 1965 theory which, in turn, influenced the interpretation of the 5. Moss L: What Genes Can’t Do. Cambridge, Mass, MIT Press, 2003 concept of a gene in the work of Mendel, in medical genet- 6. Mayr E: The Growth of Biological Thought. Cambridge, Mass, ics, and most recently in human molecular genetics. Five Belknap Press, 1982 criteria were proposed for evaluating whether the prefor- 7. Jacob F: The Logic of Life: A History of Heredity. New York, Pan- mationist concept of “X is a gene for Y” is appropriate for theon Books, 1973

1250 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KENNETH S. KENDLER

8. Falk R: The struggle of genetics for independence. J Hist Biol 28. Owen MJ, Williams NM, O’Donovan MC: The molecular genetics 1995; 28:219–246 of schizophrenia: new findings promise new insights. Mol Psy- 9. Allen GE: Mendel and modern genetics: the legacy for today. chiatry 2004; 9:14–27 Endeavour 2003; 27:63–68 29. Schwab SG, Knapp M, Mondabon S, Hallmayer J, Borrmann- 10. Davenport CB: The Feebly Inhibited: Nomadism, or the Wan- Hassenbach M, Albus M, Lerer B, Rietschel M, Trixler M, Maier dering Impulse With Special Reference to Heredity, Inheritance W, Wildenauer DB: Support for association of schizophrenia of Temperament. Washington, DC, Carnegie Institution of with genetic variation in the 6p22.3 gene, dysbindin, in sib- Washington, 1915 pair families with linkage and in an additional sample of triad 11. Rosanoff AJ, Orr FI: A Study of Insanity in the Light of the Men- families. Am J Hum Genet 2003; 72:185–190 delian Theory: Bulletin 5. Cold Spring Harbor, NY, Eugenics 30. Tang JX, Zhou J, Fan JB, Li XW, Shi YY, Gu NF, Feng GY, Xing YL, Records Office, 1911 Shi JG, He L: Family-based association study of DTNBP1 in 12. McKusick VA: Mendelian Inheritance in Man: A Catalog of Hu- 6p22.3 and schizophrenia. Mol Psychiatry 2003; 8:717–718 man Genes and Genetic Disorders, 12th ed, vols 1–3. Balti- 31. Williams NM, Preece A, Morris DW, Spurlock G, Bray NJ, more, Johns Hopkins University Press, 1998 Stephens M, Norton N, Williams H, Clement M, Dwyer S, Curran 13. Wilson EO: Sociobiology: The New Synthesis. Cambridge, Mass, C, Wilkinson J, Moskvina V, Waddington JL, Gill M, Corvin AP, Belknap Press of Harvard University Press, 1975 Zammit S, Kirov G, Owen MJ, O’Donovan MC: Identification in 2 independent samples of a novel schizophrenia risk haplotype 14. Dawkins R: The Selfish Gene. New York, Oxford University Press, 1976 of the dystrobrevin binding protein gene (DTNBP1). Arch Gen Psychiatry 2004; 61:336–344 15. Hill AB: The environment and disease: association or causa- 32. Van Den Bogaert A, Schumacher J, Schulze TG, Otte AC, Ohl- tion? Proc R Soc Med 1965; 58:295–300 raun S, Kovalenko S, Becker T, Freudenberg J, Jonsson EG, Mat- 16. Kendler KS, O’Neill FA, Burke J, Murphy B, Duke F, Straub RE, tila-Evenden M, Sedvall GC, Czerski PM, Kapelski P, Hauser J, Shinkwin R, Ni Nuallain M, MacLean CJ, Walsh D: Irish Study of Maier W, Rietschel M, Propping P, Nothen MM, Cichon S: The High-Density Schizophrenia Families: field methods and power DTNBP1 (dysbindin) gene contributes to schizophrenia, de- to detect linkage. Am J Med Genet Neuropsychiatr Genet 1996; pending on family history of the disease. Am J Hum Genet 67:179–190 2003; 73:1438–1443 17. Khuder SA: Effect of cigarette smoking on major histological 33. Kirov G, Ivanov D, Williams NM, Preece A, Nikolov I, Milev R, types of lung cancer: a meta-analysis. Lung Cancer 2001; 31: Koleva S, Dimitrova A, Toncheva D, O’Donovan MC, Owen MJ: 139–148 Strong evidence for association between the dystrobrevin 18. Agudo A, Gonzalez CA, Bleda MJ, Ramirez J, Hernandez S, Lo- binding protein 1 gene (DTNBP1) and schizophrenia in 488 pez F, Calleja A, Panades R, Turuguet D, Escolar A, Beltran M, parent-offspring trios from Bulgaria. Biol Psychiatry 2004; 55: Gonzalez-Moya JE: Occupation and risk of malignant pleural 971–975 mesothelioma: a case-control study in Spain. Am J Ind Med 34. Kendler KS, Prescott CA, Myers J, Neale MC: The structure of ge- 2000; 37:159–168 netic and environmental risk factors for common psychiatric 19. Kendler KS, Kessler RC, Walters EE, MacLean C, Neale MC, and substance use disorders in men and women. Arch Gen Heath AC, Eaves LJ: Stressful life events, genetic liability, and Psychiatry 2003; 60:929–937 onset of an episode of major depression in women. Am J Psy- 35. Kendler KS, Jacobson KC, Prescott CA, Neale MC: Specificity of chiatry 1995; 152:833–842 genetic and environmental risk factors for use and abuse/de- 20. Farrer LA, Cupples LA, Haines JL, Hyman B, Kukull WA, Mayeux pendence of cannabis, cocaine, hallucinogens, sedatives, stim- R, Myers RH, Pericak-Vance MA, Risch N, van Duijn CM (APOE ulants, and opiates in male twins. Am J Psychiatry 2003; 160: and Alzheimer Disease Meta Analysis Consortium): Effects of 687–695 age, sex, and ethnicity on the association between apolipopro- 36. Slutske WS, Eisen S, True WR, Lyons MJ, Goldberg J, Tsuang M: tein E genotype and Alzheimer disease: a meta-analysis. JAMA Common genetic vulnerability for pathological gambling and 1997; 278:1349–1356 alcohol dependence in men. Arch Gen Psychiatry 2000; 57: 21. Dick DM, Foroud T: Candidate genes for alcohol dependence: 666–673 a review of genetic evidence from human studies. Alcohol Clin 37. Scherer JF, True WR, Xian H, Lyons MJ, Eisen SA, Goldberg J, Lin Exp Res 2003; 27:868–879 N, Tsuang MT: Evidence for genetic influences common and 22. Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN: specific to symptoms of generalized anxiety and panic. J Affec- Meta-analysis of genetic association studies supports a contri- tive Disord 2000; 57:25–35 bution of common variants to susceptibility to common dis- 38. Berrettini W: Review of bipolar molecular linkage and associa- ease. Nat Genet 2003; 33:177–182 tion studies. Curr Psychiatry Rep 2002; 4:124–129 23. Ioannidis JP, Trikalinos TA, Ntzani EE, Contopoulos-Ioannidis 39. Ueno S: Genetic polymorphisms of serotonin and dopamine DG: Genetic associations in large versus small studies: an em- transporters in mental disorders. J Med Invest 2003; 50:25–31 pirical assessment. Lancet 2003; 361:567–571 40. Noble EP: D2 dopamine receptor gene in psychiatric and neu- 24. Sullivan PF, Eaves LJ, Kendler KS, Neale MC: Genetic case-con- rologic disorders and its phenotypes. Am J Med Genet B Neu- trol association studies in neuropsychiatry. Arch Gen Psychiatry ropsychiatr Genet 2003; 116:103–125 2001; 58:1015–1024 41. Zuckerman M: Behavioral Expressions and Biosocial Bases of 25. Jonsson EG, Sedvall GC, Nothen MM, Cichon S: Dopamine D4 Sensation Seeking. New York, Cambridge University Press, receptor gene (DRD4) variants and schizophrenia: meta-analy- 1994 ses. Schizophr Res 2003; 61:111–119 42. Lewis CM, Levinson DF, Wise LH, Delisi LE, Straub RE, Hovatta I, 26. Anguelova M, Benkelfat C, Turecki G: A systematic review of as- Williams NM, Schwab SG, Pulver AE, Faraone SV, Brzustowicz sociation studies investigating genes coding for serotonin re- LM, Kaufmann CA, Garver DL, Gurling HM, Lindholm E, Coon H, ceptors and the serotonin transporter, I: affective disorders. Moises HW, Byerley W, Shaw SH, Mesen A, Sherrington R, Mol Psychiatry 2003; 8:574–591 O’Neill FA, Walsh D, Kendler KS, Ekelund J, Paunio T, Lonnqvist 27. Maher BS, Marazita ML, Ferrell RE, Vanyukov MM: Dopamine J, Peltonen L, O’Donovan MC, Owen MJ, Wildenauer DB, Maier system genes and attention deficit hyperactivity disorder: a W, Nestadt G, Blouin JL, Antonarakis SE, Mowry BJ, Silverman meta-analysis. Psychiatr Genet 2002; 12:207–215 JM, Crowe RR, Cloninger CR, Tsuang MT, Malaspina D, Harkavy-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1251 GENE ACTION IN PSYCHIATRIC DISORDERS

Friedman JM, Svrakic DM, Bassett AS, Holcomb J, Kalsi G, Mc- tion in schizophrenia-spectrum disorder: long-term follow-up Quillin A, Brynjolfson J, Sigmundsson T, Petursson H, Jazin E, study of Finnish adoptees. Br J Psychiatry 2004; 184:216–222 Zoega T, Helgason T: Genome scan meta-analysis of schizo- 50. Cadoret RJ, Yates WR, Troughton E, Woodworth G, Stewart MA: phrenia and bipolar disorder, part II: schizophrenia. Am J Hum Gene-environment interaction in genesis of aggressivity and Genet 2003; 73:34–48 conduct disorders. Arch Gen Psychiatry 1995; 52:916–924 43. Mathews CA, Reus VI: Genetic linkage in bipolar disorder. CNS 51. Kendler KS: Twin studies of psychiatric illness: an update. Arch Spect 2003; 8:891–904 Gen Psychiatry 2001; 58:1005–1014 44. Tennant C: Life events, stress and depression: a review of re- 52. Caspi A, McClay J, Moffitt TE, Mill J, Martin J, Craig IW, Taylor A, cent findings. Aust NZ J Psychiatry 2002; 36:173–182 Poulton R: Role of genotype in the cycle of violence in mal- treated children. Science 2002; 297:851–854 45. Finlay-Jones R, Brown GW: Types of stressful life events and the 53. Caspi A, Sugden K, Moffitt TE, Taylor A, Craig IW, Harrington H, onset of anxiety and depressive disorders. Psychol Med 1981; McClay J, Mill J, Martin J, Braithwaite A, Poulton R: Influence of 11:803–815 life stress on depression: moderation by a polymorphism in 46. Kendler KS, Karkowski L, Prescott CA: Stressful life events and the 5-HTT gene. Science 2003; 301:386–389 major depression: risk period, long-term contextual threat and 54. Foley DL, Eaves LJ, Wormley B, Silberg JL, Maes HH, Kuhn J, Ri- diagnostic specificity. J Nerv Ment Dis 1998; 186:661–669 ley B: Childhood adversity, monoamine oxidase A genotype, 47. Kendler KS, Heath AC, Martin NG, Eaves LJ: Symptoms of anxi- and risk for conduct disorder. Arch Gen Psychiatry 2004; 61: ety and symptoms of depression: same genes, different envi- 738–744 ronments? Arch Gen Psychiatry 1987; 44:451–457 55. Risch N: Genetic linkage and complex diseases, with special 48. Eaves L, Silberg J, Erkanli A: Resolving multiple epigenetic path- reference to psychiatric disorders. Genet Epidemiol 1990; 7:3– ways to adolescent depression. J Child Psychol Psychiatry 2003; 16 44:1006–1014 56. Sarkar S: Genetics and Reductionism. New York, Cambridge 49. Tienari P, Wynne LC, Sorri A, Lahti I, Laksy K, Moring J, Naarala University Press, 1998 M, Nieminen P, Wahlberg KE: Genotype-environment interac- 57. Alfred J: Tuning in to perfect pitch. Nat Rev Genet 2000; 1:3

1252 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Images in Neuroscience Carol A. Tamminga, M.D., Editor

The Cerebellum

Molecular layer Molecular layer

Purkinje cell layer Purkinje cell layer Granule cell layer

Granule cell layer

The cerebellum means the “little brain” (when trans- brain stem that send axons called mossy fibers into the lated from Latin) and is nestled behind the “big brain” (i.e., cerebellar cortex. The mossy fibers synapse on the gran- the cerebrum) within the skull. The cerebellum tradition- ule cells and the climbing fibers from olivary neurons ally was considered to be responsible primarily for the co- wrap around the Purkinje cell dendrites as they climb into ordination of movement and motor learning. However, the molecular layer. The Purkinje cells are the only direct more recent anatomical and functional studies indicate output neurons from the cerebellar cortex, sending inhib- that the cerebellum (especially in humans) plays a wider itory projections primarily to the deep cerebellar nuclei role in many cognitive functions, such as language, execu- and a few extracerebellar regions, including the vestibular tive functions, and spatial cognition. In addition, a num- nuclei. Through indirect pathways, the cerebellum re- ber of neurologic and psychiatric conditions have been as- ceives information from all sensory modalities, including sociated with cerebellar dysfunction, including autism, auditory, visual, somatosensory, and proprioceptive sys- attention deficit hyperactivity disorder, mood disorders, tems as well as input from the neocortex. In turn, the cer- and schizophrenia. In humans, the cerebellum is a highly ebellum sends information indirectly throughout the convoluted structure; in mice its gross structure is some- brain. One important projection pathway leads to the what less complex (central part of Figure). However, in thalamus and from there to the motor cortex. Functional both humans and mice, the cellular organization is rela- neuroimaging studies in humans and physiological stud- tively simple. The cortex of the cerebellum is composed of ies in rats and mice are revealing new insights into cere- three layers of neurons: the molecular layer, the Purkinje bellar function. One thing that is becoming clear is that cell layer, and the granule cell layer. The cell bodies of the the cerebellum plays a more complex role in how the brain Purkinje neurons (in the Purkinje cell layer) extend long, functions than previously thought. elaborate, graceful dendrites into the molecular layer, as shown in the Golgi-filled Purkinje cell on the right. The major inputs into the cerebellum come from the inferior MICHAEL VOGEL, PH.D. olive neurons and from neurons in the spinal cord and Baltimore, Md.

Address reprint requests to Dr. Tamminga, UT Southwestern Medical Center, Department of Psychiatry, 5323 Harry Hines Blvd., #NC5.914, Dallas, TX 75390-9070; [email protected] (e-mail).

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1253 Introspections

In Sickness and Health

A s a physician I’ve learned a great deal about the “doctor-patient” relationship, but only a fraction of what I’ve learned when I’m the one on the other side—the one in pain. That’s when I appreciate the exquisite sense of loss of control that patients experience. That is when I begin to understand the huge power that the healers, those “health care professionals,” hold over the ones in need of help. It’s pretty easy to get it when you’re lying on a stretcher waiting for the ride back from the X-ray department to the emergency room and the technicians are talking about their lunch break or their weekend plans and you’re looking at the ceiling tiles and wondering if the pain means something really serious—something that will change your entire life. It is also pretty clear when you’re the patient and the doctors come in for rounds and make a few statements and leave, and you wonder for the next 24 hours about the implications of those few words. When I was sick this last time I had a new glimpse into that netherworld of the doctor- patient relationship. This time it wasn’t serious and I wasn’t in a hospital, just wracked with an intestinal virus that emptied my body of what seemed like all its fluids and strength. I was so depleted that my husband, who is also a physician, administered intravenous fluids to me at home over the course of an evening. Pretty unique actually, in an age when house calls by doctors have become so rare. The part about lying in my own bed next to a “The other part—the makeshift I.V. pole and using a glass jar from the husband and wife as kitchen for a “sharps container” as the fluid and medications were dripping into my arm was certainly a doctor and patient— new and interesting esthetic experience, but the was truly remarkable.” other part—the husband and wife as doctor and patient—was truly remarkable. In our 20 years together we have seen each other sick, stressed, and vulnerable, and there have been occasions to care for one another. Physical and emotional intimacy are part of our relationship, with a reassuring ebb and flow of intensity. Nothing quite prepared me, however, for the psychic shock of watching my husband survey my arm for the purpose of finding the best vein to puncture. I was lying in bed—our bed—as he loomed over me. He held me with a tender and firm grip and his touch was gentle, but his mind was more distant—somewhere else—like a doctor. He completed his preparations, then penetrated the skin and the tissue below with a sharp needle. The act he has performed so many times on others but never on— or with—me. He secured the tubing and taped it in place, tidied up, and arranged my arm on a pillow, watching satisfied as the restorative fluid began to drip into my body. “You were so nice to me,” I said in a slightly pathetic way, trying to convey the sense of appreciation of a helpless victim who is treated in a surprisingly kind manner by the often-cold doctor. “I treated you like I treat any other patient,” he replied. Strange answer. I guess I am glad that he is so nice to his patients, but maybe jealous too. On the other hand, I sensed that he was a bit distant with me—the way he needed to be to poke a hole in me with a sharp object—so I suppose he’s that way with them too. I could see and feel in a completely new way why patients fall in love with their doctors. The surge of grati- tude and admiration that arises when someone in power focuses all his attention on you—on your health and well-being, on finding a way to minimize the pain of a helpful procedure. This is the unique physical and emotional intimacy of the doctor-patient relationship, which has the potential to be misconstrued as love, exploited by those in power, and cherished as an element of healing. JEANNE STEINER, D.O. Address correspondence and reprint requests to Dr. Steiner, Connecticut Mental Health Center, Yale School of Medicine, Rm. 148, 34 Park St., New Haven, CT 06508; [email protected] (e-mail).

1254 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Images in Psychiatry

Milton Hyland Erickson, 1901–1980

Milton H. Erickson, M.D., long considered the father of such therapeutic techniques as paradox, humor, reframing, con- modern clinical hypnosis, is best appreciated today as a psycho- fusion, surprise, binds and double binds, metaphors and story- therapy innovator (1). His “uncommon” therapy (2) was molded by telling, ambiguous function assignments, variable session length, his early career research into the nature of suggestion, hypnotic and going with the resistance. Among the many hypnotic tech- states, the mental mechanisms underlying psychodynamic pro- niques he invented or developed are arm levitation, interspersal, cesses, and the psychophysiological aspects of trance (3). Dr. Erick- confusion, illogic and irrelevance, indirection, many varieties of son had a pivotal realization that even as a hypnotherapist, he could time distortion, age regression and progression, future projec- be most effective when not using formal or directive hypnosis. The tion, and trance induction by touch alone (1). nondirective, naturalistic style he invented is called Ericksonian Dr. Erickson was a truly American character, born in Nevada, hypnosis, and his revolutionary psychotherapeutic approach is who came east in a covered wagon and grew up on a farm in Wis- called Ericksonian psychotherapy. He conceptualized what he was consin. His struggles to overcome a nearly fatal bout of polio and, doing as actively catalyzing some new possibility, not as passively later, to live with postpolio syndrome shaped his therapeutic awaiting change or as commanding, prescribing, or controlling the drive. Perhaps it is not surprising that this pioneering clinician ultimate outcome. When I queried him about this in July 1977, he who spent the last 30 years of his life practicing and teaching in said, “Once you start a snowball rolling from the top of a mountain, Arizona and who used stories and metaphors frequently in his who can tell what it will grow into and what path it will take?” work with patients from near and far should himself be the sub- Dr. Erickson believed that treatment should be specifically con- ject of tales that sometimes portray him as larger than life. structed for each patient because each patient is unique. He advo- Today, nearly 100 Erickson institutes exist in almost 30 coun- cated a “utilization” approach whereby a clinician utilizes whatever tries. The American Society of Clinical Hypnosis, which Dr. Erick- behavior, ideas, or attitudes patients exhibit. Dr. Erickson had a son cofounded along with its journal, is approaching its golden deeply humane view of patients, and he spent a great deal of time anniversary. So, too, is the decision by the American Medical As- getting to know them while testing and working with their re- sociation to ratify hypnosis as a legitimate medical technique. Dr. sponses, especially their strengths. He would concoct an ap- Erickson was a courageous clinician whose ingenious insights proach—not always using hypnosis—that would allow some slight and techniques can continue to enrich our therapeutic palette if change to occur, and this was often followed by a cascade of pro- only we are willing to listen to his voice. gressive changes. He appeared to have uncanny intuition yet attributed his clinical insight to both the acute development of his own References perceptual skills and the innate receptive and synthetic capacities 1. Zeig JK, Munion WM: Milton H Erickson. Thousand Oaks, Calif, of the unconscious, which Dr. Erickson viewed as a reservoir of cre- Sage Publications, 1999 ative potential that can be a source of wisdom, not just of pathology. 2. Haley J: Uncommon Therapy: The Psychiatric Techniques of Among the many therapeutic approaches he spawned were Milton H Erickson, MD. New York, WW Norton, 1973 one-session therapy, brief therapy, strategic family therapy, sys- 3. Erickson MH: The Collected Papers of Milton H Erickson on Hyp- tems-oriented therapy, solution-focused therapy, ordeal therapy, nosis, vols 1–4. Edited by Rossi EL. New York, Irvington, 1980 neurolinguistic programming, pediatric and dental hypnoanal- gesia, and Ernst Rossi’s psychobiological therapy. He invented GREGG E. GORTON, M.D.

Address reprint requests to Dr. Gorton, Philadelphia VA Medical Center, Department of Psychiatry, Mailstop 116, University and Woodland Aves., Philadelphia, PA 19140; [email protected] (e-mail). Photograph provided with permission of the Archives of the Milton Erickson Foundation.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1255 Article

Structural Brain Magnetic Resonance Imaging of Limbic and Thalamic Volumes in Pediatric Bipolar Disorder

Jean A. Frazier, M.D. Background: Youths with bipolar disor- of variance, with total cerebral volume and age controlled. Sufen Chiu, M.D., Ph.D. der are ideal for studying illness pathophysiology given their early presentation, lack Results: The subjects with bipolar disor- Janis L. Breeze, M.P.H. of extended treatment, and high genetic der had smaller hippocampal volumes. Nikos Makris, M.D., Ph.D. loading. Adult bipolar disorder MRI studies Further analysis revealed that this effect Nicholas Lange, Sc.D. have focused increasingly on limbic struc- was driven predominantly by the female tures and the thalamus because of their role bipolar disorder subjects. In addition, David N. Kennedy, Ph.D. in mood and cognition. On the basis of both male and female youths with bipo- Martha R. Herbert, M.D., Ph.D. adult studies, the authors hypothesized a lar disorder had significantly smaller cere- priori that youths with bipolar disorder bral volumes. No significant hemispheric Eileen K. Bent, B.A. effects were seen. would have amygdalar, hippocampal, and Vamsi K. Koneru, B.A. thalamic volume abnormalities. Conclusions: These findings support the Megan E. Dieterich, B.A. hypothesis that the limbic system, in par- Method: Forty-three youths 6–16 years of ticular the hippocampus, may be in- Steven M. Hodge, M.A. age with DSM-IV bipolar disorder (23 male, volved in the pathophysiology of pediat- Scott L. Rauch, M.D. 20 female) and 20 healthy comparison ric bipolar disorder. While this report may subjects (12 male, eight female) similar in represent the largest MRI study of pediat- P. Ellen Grant, M.D. age and sex underwent structured and ric bipolar disorder to date, more work is Bruce M. Cohen, M.D., Ph.D. clinical interviews, neurological examina- needed to confirm these findings and to Larry J. Seidman, Ph.D. tion, and cognitive testing. Differences in determine if they are unique to pediatric bipolar disorder. Verne S. Caviness, M.D., D.Phil. limbic and thalamic brain volumes, on the logarithmic scale, were tested using a two- Joseph Biederman, M.D. way (diagnosis and sex) univariate analysis

(Am J Psychiatry 2005; 162:1256–1265)

Bipolar disorder is one of the most severe neuropsy- Although no discrete brain area has been consistently chiatric disorders at any age and is among the most dis- reported abnormal in the adult bipolar disorder MRI abling of psychiatric conditions that affect youths (1, 2). (structural) literature, some data lend support to several Although the chronology of underlying structural brain proposed neuroanatomic models of emotion regulation abnormalities in this population is unknown, such abnor- (8–10). One of these proposed models includes the follow- malities may represent disruptions in typical brain growth ing brain regions: prefrontal cortex, amygdala-hippo- resulting from an interplay of genetic and environmental campus complex, hypothalamus, thalamus, insular cor- factors. MRI studies are critical for advancing our knowl- tex, ventral striatum, and interconnected structures (8). The amygdala, hippocampus, and thalamus are of partic- edge of brain regions involved in the pathophysiology of ular interest in the study of bipolar disorder because of pediatric bipolar disorder. their functional roles in the brain. For example, the amyg- Youths with bipolar disorder are more severely ill and dala is integral in emotion-related aspects of behavior, have higher genetic loading than adult-onset cases (3–5). memory, and learning; the thalamus processes sensory in- Furthermore, children’s brains are typically free from con- formation and integrates activity among forebrain re- founding factors known to affect brain structure and func- gions; and the hippocampus plays a role in learning and tion (e.g., extensive treatment, substance use, history of memory in providing contextual information (11). Several electroconvulsive therapy). Finally, studying children and prior structural MRI studies of adults with bipolar disorder adolescents facilitates observation during a time in devel- have reported abnormalities in the limbic structures and opment when there are hormonal shifts known to have thalamus (12–18). Abnormalities in these structures might neuromodulatory effects on brain regions such as the confer a propensity toward dysregulated mood states and temporal lobe (6, 7). By virtue of all of these factors, there vulnerability toward developing a mood disorder. is an increased likelihood of uncovering significant brain Prior structural MRI studies in pediatric bipolar disor- anatomic abnormalities in this early-onset group. der have indicated that there are anatomic abnormalities

1256 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FRAZIER, CHIU, BREEZE, ET AL. in a number of the structures implicated in the neural sys- were determined ineligible during interview, and one stopped the tems governing affective and cognitive processes. Bot- study because of lack of interest. Sixty-six scans were obtained; teron and colleagues demonstrated a loss of the normal three scans were unreadable due to motion artifact. Therefore, data from 63 subjects scanned as part of an ongoing neuroimag- asymmetry in the frontal lobe (19). A study conducted by ing study are included in this report: 43 youths with DSM-IV bipo- Friedman and colleagues (20), which included adoles- lar disorder and 20 healthy subjects. cents with schizophrenia and bipolar disorder in a com- All of the youths underwent a diagnostic semistructured inter- bined patient group, found that this patient group had review (Schedule for Affective Disorders and Schizophrenia for duced intracranial volumes and increased frontal and School-Age Children—Epidemiologic Version [K-SADS-E] [23]) and a clinical interview by board-certified child psychiatrists temporal sulcal sizes relative to healthy subjects. A third (J.A.F., S.C.). In addition, parents were administered an indirect K- report that used the same subjects as in the Friedman et al. SADS-E regarding their children by trained raters. These B.A.- study found the patient group had reduced thalamic area level raters received 4 months of training on the administration of relative to healthy subjects (21). A recent study consisting the K-SADS-E under the supervision of senior raters and the se- predominantly of adults with bipolar disorder included a nior investigator (J.B.). All raters had established a high degree of interrater reliability: from 175 interviews, the mean kappa was subgroup of 14 adolescents (age range=10–22 years). 0.90, and all disorders achieved kappa coefficients >0.82. Final These youths had smaller hippocampal and amygdalar DSM-IV diagnoses were established by the consensus diagnosis volumes than did 23 healthy adolescents (18). Finally, Del- of clinical and structured interviews. Bello and colleagues (22) recently reported that adoles- Each youth received a physical and neurological examination cents with bipolar disorder (N=23) had smaller amygdala that included Tanner staging (a I–V scale of pubertal develop- and enlarged putamen volumes compared with healthy ment) (24) and cognitive testing. The age at onset of each illness was determined by parental report of symptoms on the struc- subjects (N=20). In summary, prior MRI studies have sug- tured interview. Age at illness onset was defined as the time when gested that youths with bipolar disorder have abnormali- the youth met full diagnostic criteria (e.g., age at onset of bipolar ties in a number of the brain areas discussed by Soares and illness was the age at which the youth first met full diagnostic cri- Mann (8) in their neuroanatomic model of emotion regu- teria for mania). Children and adolescents were given several sub- lation: total cerebral volume, frontal lobe, hippocampus, tests of the Wechsler Intelligence Scale for Children, 3rd ed. (WISC-III) (25) which permitted the estimation of verbal IQ. amygdala, putamen, and thalamus. In order to assess ana- Handedness was assessed using the Edinburgh Handedness tomic findings in bipolar disorder further, we conducted a Questionnaire (26). structural MRI study to evaluate brain volumes in early- Measures of current psychopathology were obtained using the onset bipolar disorder cases. We hypothesized that youths Young Mania Rating Scale (27) and Global Assessment of Func- with bipolar disorder would have abnormalities in struc- tioning Scale (GAF) (DSM-IV, p. 32). tures involved in the model of affect regulation discussed Antipsychotic doses (converted to chlorpromazine equiva- by Soares and Mann (8). In particular, the amygdala, lents) (28, 29), as well as number and type (antipsychotic, antidepressant, stimulant, anticonvulsant, lithium) of psychoactive hippocampus, and thalamus were chosen a priori on the medications at the time of scan were used as clinical variables. basis of previous imaging studies of youths and adults with bipolar disorder. MRI Protocol Structural imaging was performed at the McLean Hospital Brain Method Imaging Center on a 1.5-T Signa scanner (GE Medical Systems, Milwaukee). Acquisitions included a conventional T1-weighted Subjects sagittal scout series (20 slices), a proton density/T2-weighted in- The study was approved by institutional review boards at the terleaved double-echo axial series (120 slices, slice thickness=3 2 Massachusetts General Hospital and McLean Hospital. Subjects mm, field of view=24 cm , TR=3 seconds, TE=30/80 msec, acqui- × were recruited through the McLean Hospital outpatient program sition matrix=256 192, number of excitations=0.5), and a three- and professional-patient advocacy groups. Inclusion criteria were dimensional inversion recovery-prepped spoiled gradient re- DSM-IV diagnosis of bipolar disorder, age 6–16 years, and right- called echo coronal series, which was used for structural analysis ° handedness. Male and female subjects of all ethnicities were re- (124 slices, prep=300 msec, TE=1 minute, flip angle=25 , field of 2 cruited. Healthy subjects, all right-handed, were recruited through view=24 cm , slice thickness=1.5 mm, acquisition matrix= × community newspaper advertisements and had no DSM-IV axis I 256 192, number of excitations=2). All scans were reviewed by a diagnosis according to structured and clinical interviews and no clinical neuroradiologist to rule out gross pathology. family history of affective disorders or psychotic disorders in first- Image Analysis degree relatives. Exclusion criteria were major sensorimotor hand- icaps; full-scale IQ <70 or learning disabilities; history of claustro- Structural scans were transferred to the NMR Center for Mor- phobia, head trauma, loss of consciousness, autism, schizophre- phometric Analysis-Charlestown Massachusetts General Hospi- nia, anorexia or bulimia nervosa, electroconvulsive therapy, or tal and coded and catalogued for blind analysis. Imaging analysis alcohol or drug dependence/abuse (in the 2 months preceding the was done on Sun Microsystems, Inc. (Mountainview, Calif.) work- scan or a total history of 12 or more months); active medical or stations using Cardviews software (30). The datasets were posi- neurologic disease; metal fragments or implants; or current preg- tionally normalized to overcome variations in head position by nancy or lactation. imposing a standard orientation on each scan using the midpoints of the decussations of the anterior and posterior commissure Procedure lines and the midsagittal plane at the level of the posterior com- Seventy subjects (all outpatients) and their parents (or guard- missure as points of reference for rotation and translation. The ians) signed assent and informed consent forms. Three subjects images were not rescaled to Talairach spatial dimensions in order

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1257 NEUROIMAGING FINDINGS IN PEDIATRIC BIPOLAR DISORDER

FIGURE 1. T1-MRI Coronal Section at the Level of the B, C); this preliminary procedural step allows a reliable separation Mamillary Bodiesa of the amygdala from surrounding gray structures, such as the ventral part of the lentiform nucleus, the medial temporal cortex, and the hippocampus, thus eliminating the need to apply conventions to define the anterior amygdalar boundary (12, 22) or Cortex the amygdala-hippocampal junction (12). The anterior portion of White the amygdala was segmented as it appears beneath the medial Matter temporal cortex (slice K in Figure 2). At this region, the medial temporal cortex and the amygdala can give the impression of be- Lateral Ventricle ing only a thickening of the medial temporal cortex with no amyg- Caudate dala present, as has been reported previously (12, 22). Therefore, the definition of these borders was particularly aided by the trac- Thalamus ing of cross-referenced outlines in the axial and sagittal planes Putamen (Figure 2 B, C). The choroidal fissure was used as the superior bor- Pallidum der of the amygdala along with the gray-white matter contrast between the amygdala and surrounding white matter. The gray- Amygdala Hippocampus white matter contrast between the amygdala and its surrounding temporal white matter (consisting of the centrally located temporal white matter stem), as well as the gray-CSF contrast between the amygdala and the temporal horn of the lateral ventricle, was considered the lateral border of the amygdala. The parahippocampal cortex anteriorly, the brain exterior at the inferior lip of the choroidal fissure, and partially the hippocampus posteriorly a Segmentation outlines shown in green. were assigned as the medial borders of the amygdala. Finally, the inferior border consisted of the gray-white matter contrast between the amygdala and its surrounding temporal white matter to preserve individual and interhemispheric differences in the morphometry of structures (31). anteriorly and by the alveus (of the hippocampus) and the temporal horn of the lateral ventricle posteriorly. The entire image sets were then segmented into gray, white, and CSF tissue classes by three image analysts, under the supervision of The volumes for each structure were derived by multiplying the one of the authors (N.M.), all of whom had strong backgrounds in number of voxels assigned to each structure on each slice by the neuroanatomy and extensive training in morphometric analysis voxel volume (the product of slice thickness and the square of in- and who were blind to subject-identifying information. The seg- plane resolution), followed by summing across all slices in which 3 mentation method uses a semiautomated intensity contour algo- the structure appeared; the volumes are reported in cm (31). rithm for external border definition and signal intensity histogram For the reliability study, 10 scans from our data set were se- distributions for delineation of gray-white borders (Figure 1). This lected at random and blindly segmented by two raters. Five of the technique allows for border definition as the midpoint between the scans were also remeasured in a random order by one of the raters peaks of the bimodal distribution for any given structure and its to estimate the intraclass correlation coefficient. surrounding tissue (32). The standard interrater intraclass correlation coefficient for the total cerebrum was 0.93, and the intra- and interrater correla- Total cerebral volume. Segmentation of the regions of interest tion coefficients, respectively, for the regions of interest were 0.88 was performed following the anatomic definitions of Filipek and and 0.84 for the amygdala, 0.95 and 0.96 for the thalamus, and colleagues (33) for the total cerebrum. Total cerebral volume was 0.93 and 0.94 for the hippocampus. defined as all gray and white matter in the cerebrum and did not include CSF, cerebellum, or brain stem. The cerebrum was mea- Data Analyses sured across all 124 coronal slices in which it appeared. S-Plus 6.0 (Insightful Corp., Seattle) was used for statistical Thalamus. Segmentation of the thalamus was performed fol- analysis. All statistical tests were two-sided with alpha set at 0.05. lowing the anatomic definitions of Seidman and colleagues (34, Differences in demographic and clinical variables were mea- 35) by tracing the trajectory of the hypothalamic fissure in the sured using t tests for continuous variables and chi-square tests sagittal plane to separate the thalamus proper from the ventral di- for categorical variables. In addition, Pearson’s correlations were encephalon. The medial boundary of the structure was the third calculated for clinical variables of the bipolar disorder group and ventricle, and the lateral boundary was the internal capsule. The those structures that differed significantly between bipolar disor- superior border was the body of the lateral ventricle, and the infe- der youths and healthy subjects. The clinical variables included rior border was the hypothalamic fissure. Young Mania Rating Scale and GAF scores, number of psycho- Hippocampus and amygdala. The method of Filipek et al. active medications, chlorpromazine equivalents, and verbal IQ. (33) defines the amygdala and hippocampus as a continuous gray Noting that brain structure sizes possess variances that increase matter structure in the primary segmentation. These two struc- with their means, we analyzed our volumetric data on the natural tures are then separated from each other according to the proce- logarithmic scale as a step toward uncoupling this non-Gaussian dure described by Seidman and colleagues (36) in which the relationship. We conducted an exploratory multivariate analysis of hippocampus is separated from the amygdala at the rostral-coro- variance (MANOVA) on the three-dimensional ensemble of log nal plane, where the hippocampus first appears. The segmenta- total hippocampal, amygdalar, and thalamic volumes with the tion of the amygdala was performed manually in its entirety, effects of log total cerebral volume, age, sex, diagnosis, and the comprising approximately 11 subsequent coronal sections. sex-by-diagnosis interaction controlled. In addition, an explor- The coexistence of the amygdala and hippocampus in several atory MANOVA analysis was performed to assess the effects of age coronal sections can make the precise identification of the ventral group, mood state, medication type, and presence of ADHD or amygdalar border difficult (37). Therefore, we used the cross-ref- psychosis on the log hippocampal and log total cerebral volumes. erencing capability of the program Cardviews (38) to draw out- We then proceeded to analyze the effects of these covariates on lines delimiting the amygdala in axial and sagittal views (Figure 2 log brain structure sizes by three separate univariate linear regres-

1258 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FRAZIER, CHIU, BREEZE, ET AL.

FIGURE 2. Segmentation of the Amygdala, Hippocampus, and Thalamusa

A BC Amygdala Amygdala

Hippocampus Hippocampus JKLMNOPQ

DEF

G HI Thalamus Thalamus Thalamus Amygdala Amygdala

Hippocampus Hippocampus

Amygdala

LMAmygdala

Amygdala Parahippocampal gyrus Hippocampus

NOAmygdala

Hippocampus Hippocampus

Hippocampus a The segmentation method of the amygdala, hippocampus, and thalamus is shown in T1-weighted MR images. In particular, the segmentation method used for the amygdala (developed by N.M.) is shown in detail in images J–P. Image A shows a sagittal slice passing through the hippocampus and amygdala. Yellow vertical lines show the locations of representative coronal sections J–Q. The blue box in A indicates the region containing amygdala (K–N), amygdala-hippocampal transition, anterior hippocampus (L), and the temporal-polar region rostral to the amygdala (J). B shows an axial section (the position of which is indicated by arrow in A) used for the tracing of an outline to separate the most anterior portion of the amygdala from the surrounding medial temporal cortex (red line). C shows an enlarged version of the blue box in A to emphasize the amygdala-hippocampal transition. In both B and C, the light green line represents the border between amygdala and hippocampus. In C, the red line shows the superior, anterior, and inferior limits of the amygdala. These outlines are used as guidelines to complete segmentation on the coronal plane (D–I). Similarly, the thalamus is segmented based on the intensity contrast between this structure and its surrounding white matter; a border delimits its inferior border at the hypothalamic fissure. Structural details from coronal slices J–O are enlarged to show the method of amygdala segmentation. J shows a rostral slice in the temporal lobe where the amygdala is not present. More caudally, the anterior part of the amygdala appears within the rostromedial temporal area beneath the cortex as shown in K. Further posteriorly, the middle portion of the amygdala occupies a position superior and lateral to the anterior parahippocampal gyrus as shown in L. Progressing in the rostrocaudal dimension, the hippocampus appears as shown in M and N, in which the amygdala is also present. In a more posterior location (O), there is only hippocampus as amygdala is no longer present. In P, the posterior most segment of the hippocampus is shown flanked laterally by the atrium of the lateral ventricle. Finally, Q is the coronal posterior to the hippocampus. To emphasize the amygdala and hippocampal outlines, the segmentation outlines of the other brain structures were omitted from images J–Q.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1259 NEUROIMAGING FINDINGS IN PEDIATRIC BIPOLAR DISORDER

TABLE 1. Demographic and Clinical Characteristics of Pedi- (19% [N=8]), which included anticholinergics and beta- atric Patients With Bipolar Disorder and Healthy Compari- adrenergics. Table 2 presents details of the clinical and son Subjects treatment characteristics of the bipolar disorder patients. Bipolar Patients Healthy Subjects Characteristic (N=43) (N=20) Clinical neuroradiological interpretations of the scans N % N % showed normal variants in three healthy subjects (slightly Sex prominent lateral ventricles [N=1], large cisterna magna Male 23 53.5 12 60.0 [N=1], and pineal cyst [N=1]) and two bipolar disorder Female 20 46.5 8 40.0 Caucasian 4297.71890.0subjects (mildly prominent lateral ventricles [N=1] and large cisterna magna [N=1]). One healthy subject had find- Mean SD Mean SD ings that were atypical but of unclear clinical significance (a tiny focus of hypointensity in the subcortical white mat- Age (years) 11.3 2.7 11.0 2.6 Handedness score ter of the superior left frontal lobe). Nine bipolar disorder (right minus left) 15.7 9.3 16.9 5.1 subjects had findings that were atypical and again of un- Height (cm) 147.1 14.2 142.2 15.2 clear clinical significance (prominent ventricles [N=3], Weight (kg)a 49.6 14.2 41.1 15.5 Head circumference (cm) 54.3 1.4 54.8 1.8 right-greater-than-left asymmetry of the temporal horn Tanner stage 2.5 1.7 2.0 1.6 [N=2], multiple white matter hyperintensities in the bilat- Hollingshead socioeconomic eral parietal area [N=1] and in the left hemisphere [N=1], status rating 48.9 13.0 54.4 17.2 nonspecific punctate T foci in the left parietal region [N= a Significant between-group difference (t=–2.0, df=61, p=0.05). 2 1], and bilateral widened perivascular spaces noted in the inferolateral portion of the basal ganglia [N=1]). sion models. Employing our preliminary data summaries and linear regressions, we found no significant laterality effects on any of Volumetric Measurements the volumetric outcomes; hence, we did not include brain hemisphere effects in our model selection procedure. We fit a linear re- Table 3 contains the limbic and thalamic volume data. gression model for log total hippocampal volume that contained The amygdala volumes were measured across a mean of the same effects as in the preliminary MANOVA. We also analyzed 12.1 slices (SD=1.6, range=9–15) on the right side, and 11.5 the data using the absolute volumes of the structures and the in- slices (SD=1.4, range=8–15) on the left. The hippocampal ferential findings were identical to the log scale analyses. volumes were measured across a mean of 24.4 slices (SD= 1.7, range=21–28) on the right side, and 24.7 slices (SD=2.0, Results range=20–29) on the left. The thalamic volumes were mea- Data from 63 subjects are included: 43 youths with sured across a mean of 22.7 slices (SD=1.3, range=18–25) DSM-IV bipolar disorder (mean age=11.3 years, SD=2.7; on the right side, and 22.9 slices (SD=1.1, range=21–25) on current episode: mixed=52.3%, manic=15.9%, depressed= the left. 11.4%, euthymic=20.5%) and 20 healthy comparison sub- No significant correlations were seen between any clin- jects (mean age=11.0 years, SD=2.6). Characteristics of the ical variables and the hippocampus or total cerebral vol- groups are summarized in Table 1. There were no signifi- ume for the bipolar disorder group. cant height, head circumference, or Tanner stage differ- When the bipolar disorder group was assessed using the ences between groups; there was a significant difference exploratory MANOVA analysis, we found no significant ef- in weight (Table 1). fects of age group, mood state, medication type, or pres- Although the bipolar disorder youths had verbal IQ ence of ADHD or psychosis on volume in the hippocam- scores in the normal range, they were significantly lower pus or cerebrum. than those of comparison subjects (mean=100.7 [SD=14.3] The initial exploratory MANOVA of the three-dimen- versus 116.9 [SD=12.7]; t=4.3, df=61, p<0.001). The bipolar sional brain structure ensemble yielded potentially signif- disorder group also scored lower on the GAF (mean=49.5 icant effects for sex (F=2.76, df=3, 55, p=0.05) and the sex- [SD=6.0] versus 68.8 [SD=1.9]; t=19.0, df=61, p<0.001). The by-diagnosis interaction (F=2.17, df=3, 55, p=0.10). bipolar disorder subjects had a number of comorbid con- Hippocampus. The linear regression model for log total ditions (mean=7.2, SD=3.2). The most common comorbid hippocampal volume contained the same effects as in the conditions were oppositional defiant disorder (67% [N= preliminary MANOVA. We found a possible sex-by-diag- 29]) and ADHD (51% [N=22]). Most bipolar disorder youths nosis interaction in raw hippocampal volumes by sex and (86% [N=37]) had experienced at least one episode of ma- diagnosis (Table 3). Although not significant, we retained jor depression (mean age at onset=6.8 years, SD=3.6), and age in the model, since this covariate was fixed by our 40% (N=17) had a history of psychosis. One youth had a study design; main effects and the interaction of sex and history of alcohol abuse, which occurred more than 3 diagnosis remained significant when age was dropped. We months before enrollment. Medications used at the time of tried controlling for height and weight but found that nei- MRI included lithium (26% [N=11]), anticonvulsants (42% ther of their effects was significant; the sex-by-diagnosis [N=18]), antidepressants (30% [N=13]), stimulants (21% interaction remained significant. We found significant ef- [N=9]), atypical antipsychotics (76% [N=33]), and other fects of log cerebral volume, sex, diagnosis, and the sex-

1260 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FRAZIER, CHIU, BREEZE, ET AL.

TABLE 2. Clinical and Treatment Characteristics of Pediatric Patients With Bipolar Disorder Pediatric Bipolar Patient Group Characteristic Total (N=43) Boys (N=23) Girls (N=20) Mean SD Mean SD Mean SD

Age at onset of bipolar disorder (years) 7.0 3.8 5.8 3.6 8.8 3.8 Age at onset of ADHDa 4.82.44.32.36.02.4 Duration of bipolar disorder (years) 2.8 3.1 2.8 2.7 2.9 4.2 Young Mania Rating Scale Scoreb 24.8 8.0 22.6 8.2 28.0 6.9 Age at first hospitalizationc 6.45.04.54.08.65.6 Chlorpromazine equivalents at time of studyd 117.1 98.0 121.6 100.5 111.5 97.6 Number of psychoactive medications at time of studye 2.11.12.31.12.11.0

N % N % N %

History of hospitalization 14 32.6 7 30.4 7 35.0 a Total N=22. b Significant gender difference (t=2.1, df=41, p=0.04). c Total N=14. d Total N=33 (77%). e Includes atypical antipsychotics, antidepressants, sedatives, mood stabilizers, and stimulants.

TABLE 3. Limbic and Thalamic Volumes of Pediatric Patients With Bipolar Disorder and Healthy Comparison Subjects Bipolar Disorder Healthy Subjects Mean Coefficient of Mean Coefficient of Brain Area and Patient Group N (cm3)SDVariation (%) N (cm3)SDVariation (%) Amygdala Boys 23 3.48 0.52 14.9 12 3.69 0.68 18.4 Girls 20 3.16 0.36 11.4 8 3.46 0.43 12.4 Total 43 3.33 0.47 14.2 20 3.60 0.59 16.5 Hippocampus Boys 23 7.61 0.70 9.3 12 7.87 0.87 11.0 Girls 20 6.83 0.75 11.1 8 8.16 0.62 7.7 Total 43 7.24 0.82 11.3 20 7.98 0.77 9.7 Thalamus Boys 23 16.36 0.88 5.4 12 16.87 1.18 7.0 Girls 20 15.02 0.99 6.6 8 16.15 1.05 6.5 Total 43 15.74 1.14 7.3 20 16.58 1.16 7.0 by-diagnosis interaction (Table 4). We pursued the source (12, 22), and smaller amygdala (18, 22) and hippocampal of interaction by fitting separate regression models of the (18) volumes relative to healthy subjects. same form to the boys and the girls. This step confirmed These findings in children and adolescents differ some- our pooled variance assumption when fitting the single what from those in adults, but comparison of MRI results model with a sex effect (39). We conclude that the sex-by- across studies is difficult because of variations in method- diagnosis interaction is being driven by the smaller hippo- ology. For example, until recently, it was difficult to even re- campus of the girls after log total cerebral volume and age liably measure structures such as the hippocampus—and are controlled. particularly the amygdala—on MRI scans (12, 22). It is of Amygdala and thalamus. The same logarithmic trans- note that the quality of the T1-weighted images used in the formations and model selection procedure was applied to present study, as well as the outlines traced on the cross-ref- the thalamic and amygdalar volumes separately. No sig- erenced axial and sagittal slices, allow reliable visualization nificant effects of sex, diagnosis, or their interaction were and segmentation of the amygdala, eliminating the need to found. apply conventions for defining either the anterior amygdalar boundary (12, 22) or the amygdala-hippocampal Discussion junction (12). Despite the differences in the methodologies used across studies, it is worth highlighting the anatomic The youths in our study with bipolar disorder had sig- variations that have been reported in adult structural MRI nificantly smaller total hippocampal and total cerebral studies relative to the findings reported here and in other volumes. These findings contribute to the existing litera- pediatric studies of bipolar disorder patients. ture on bipolar disorder youths, which has shown loss of Of the structures included in our a priori hypothesis, the normal frontal lobe asymmetry (19), reduced intracranial thalamus and amygdala have varied findings in adult and volume (20), increased frontal and temporal sulcal size child and adolescent studies. Several adult bipolar disor- (20), reduced thalamic area (21), larger putamen volume der studies have reported that both structures are in-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1261 NEUROIMAGING FINDINGS IN PEDIATRIC BIPOLAR DISORDER

TABLE 4. Total Cerebral Volume, Age, Sex, Diagnosis, and ther decreased or normal volumes (12, 16, 18, 41). It is pos- Sex-by-Diagnosis Interaction Effects on Limbic and Thalamic sible that the adult studies in which reduced hippocampal Volumes in Pediatric Patients With Bipolar Disorder and Healthy Comparison Subjects volumes were reported may have included at least some Estimated t adults who had childhood-onset illness. However, this in- Brain Area and Covariate Effect SE (df=61) p formation was not included in those studies. If age at on- Amygdala set, particularly childhood-onset, is related to reduced hip- Log total cerebral volume 0.872 0.272 3.209 0.002 pocampal volumes, this could partially explain the mixed Age 0.002 0.007 0.229 0.82 Sex 0.019 0.064 0.299 0.77 results seen in adult studies. Our finding of reduced hip- Diagnosis –0.013 0.062 –0.206 0.84 pocampal volume, as well as another group’s similar find- Sex-by-diagnosis –0.003 0.076 –0.040 0.97 ing among adolescents with the illness (18), may represent Hippocampus Log total cerebral volume 1.068 0.150 7.134 <0.001 a finding unique to early-onset bipolar disorder. This find- Age –0.001 0.004 –0.212 0.84 ing in our youths is of interest in light of neuropathologic Sex –0.090 0.035 –2.558 0.02 studies of the hippocampus in bipolar disorder, which Diagnosis –0.086 0.034 –2.529 0.02 Sex-by-diagnosis 0.100 0.042 2.385 0.02 have suggested there may be abnormal neurodevelopment Thalamus and remodeling of synapses in that structure (42–48). Log total cerebral volume 0.593 0.097 6.143 <0.001 Age 0.005 0.003 1.832 0.08 A study of adults that compared monozygotic twins Sex 0.022 0.023 0.957 0.35 discordant for bipolar disorder found that the right hip- Diagnosis –0.024 0.022 –1.098 0.28 pocampus was smaller in the sick twin compared with the Sex-by-diagnosis 0.022 0.027 0.807 0.43 well twin (17), suggesting that this finding might be a structural correlate for the presence of disease. Given that creased (12, 13, 15), whereas two others reported smaller reduced hippocampal volume was the only significant volumes: one reported smaller left amygdala (14) and the limbic finding in our group of bipolar disorder youths, and other smaller bilateral amygdalar volumes (18). In adoles- that at least two studies in adults (16, 17) had a similar cents, one study found reduced thalamic area (although finding in the right hippocampus, this may also be a find- the patient group also included youths with schizophre- ing reflective of disease. nia) (21). Blumberg and colleagues (18) and DelBello and Female bipolar disorder subjects showed a more pro- colleagues (22) have recently reported decreased amyg- nounced decrease in hippocampal volume than did male dalar volumes in adolescents with bipolar disorder. In our bipolar youths relative to their comparison subjects. This study, youths with bipolar disorder had a nonsignificant sex-by-diagnosis interaction could reflect a variety of fac- but possibly meaningful decrement in thalamic volumes tors. For example, girls may need to have more significant compared with healthy subjects; our current study size structural abnormalities in order to reach threshold for may be underpowered to detect this effect, and we con- disease expression. Alternatively, since the girls with bipo- tinue to study this structure as more scans are acquired. lar disorder in our study had significantly higher Young However, we found that there were no differences in amyg- Mania Rating Scale scores than the boys, the smaller hip- dalar volume between groups. pocampal volumes in girls might be reflective of a greater Interpreting the differences between adult and pediatric degree of psychopathology. However, both male and fe- findings depends on our understanding of the phenome- male bipolar disorder youths had reduced hippocampal nology of mood disorders across the lifespan: if early- and volumes, and it may be that there is some abnormality in adult-onset bipolar disorder are the same illness, then age the hippocampus, reflective of disease, that expresses it- at onset may differentially affect brain structure. Alterna- self more robustly in girls than in boys (49). tively, childhood- or adolescent-onset bipolar disorder In our study we found that total cerebral volume was may be a distinct disorder from adult-onset illness with a smaller in bipolar disorder youths (mean=5.4%, SD=6.4%); different set of neuroanatomic correlates. Cross-sectional smaller total cerebral volume has also been reported in measures of anatomic volume also may not adequately four prior studies of bipolar disorder youths (19–21, 50). reflect the progressive and regressive patterns of growth Most adult bipolar disorder studies, including the recent that affect different brain structures at various stages of study published by Blumberg and colleagues (18), have not development (e.g., the subcortical gray matter structural found differences in total cerebral volume. Therefore, the changes that occur in the temporal lobe during the pre- difference seen in total cerebral volume between early- and teen years [38, 40]). adult-onset bipolar disorder cases relative to healthy Our finding of decreased hippocampal volumes is the groups suggests that affected youths may have distinct first report of this finding in children and adolescents with neurodevelopmental trajectories, possibly from having bipolar disorder and further extends the findings of Blum- brains that have developed differently or from early apop- berg and colleagues in which bilateral hippocampal vol- totic pruning of neuronal circuits. A smaller total cerebral ume reductions were seen in adolescents with bipolar dis- volume in early-onset bipolar disorder illness may be re- order but not in adults with the illness (18). Adult bipolar flective of a neurodevelopmental phenomenon. It should disorder studies of hippocampal volumes have found ei- be noted that there is evidence of nonuniform scaling

1262 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FRAZIER, CHIU, BREEZE, ET AL. across the brain regions studied herein relative to global comparison group to sort out diagnostic, age at onset, de- changes in total cerebral volume (51). velopmental, and sexual influences on these structures Although abnormal anatomy does not necessarily conover time (6, 38, 58). A larger group of subjects is currently fer abnormal function, some structural differences may being accrued in order to further assess these findings increase the risk of developing dysfunction. Such differ- with greater power and to assess the sexual dimorphism of ences might also be markers of preexisting susceptibility other brain structures. or vulnerability and could aid in identifying different patient phenotypes (52). Bipolar disorder youths often have Received June 24, 2003; revisions received Dec. 29, 2003, and July 9, 2004; accepted July 19, 2004. From the Department of Psychiatry behavioral and developmental difficulties early in life (2, and the Neuroscience Program, Harvard Medical School, Boston; the 52–54), and their symptoms of bipolar disorder typically McLean Hospital Child Psychiatry Outpatient Clinic, Brain Imaging begin between the ages of 5–11 years (1, 3, 55); in our study Center, Statistical Neuroimaging Laboratory, and Developmental Neuroscience Laboratory; the Departments of Psychiatry, Pediatric the mean age at onset was 7.0 years (SD=3.8). Many of the Neurology, Radiology, Neurology, and Pediatrics, the Pediatric Psy- progressive and regressive events in the brain, particularly chopharmacology Unit, and the Center for Morphometric Analysis, in the temporal lobe, occur during this age range, and al- Massachusetts General Hospital, Boston; the Department of Psychia- try and Behavioral Sciences, University of California, Davis; the De- terations in normal brain development during this time partment of Biostatistics, Harvard School of Public Health, Boston; may result in the symptoms of bipolar disorder. and the Department of Psychiatry, Massachusetts Mental Health Cen- ter, Boston. Address correspondence and reprint requests to Dr. Adult genetic and neuroimaging studies in both schizo- Frazier, Child and Adolescent Neuropsychiatric Research Program, phrenia and bipolar disorder lend support for a multifac- Cambridge Health Alliance/Mystic Center, 1493 Cambridge St., Cam- toral etiology; one possibility is of a “two-hit” hypothesis, a bridge, MA 02139; [email protected] (e-mail). Supported by NIMH research grants to Dr. Frazier (K08 MH-01573- genetic predisposition in combination with environmen- 01) and Dr. Lange (NS-37483). tal influences, resulting in the disorder (35, 36, 56, 57). For The authors thank Mary Ahn, M.D., Sandra DeJong, M.D., and Jay example, our finding of smaller hippocampal volumes Giedd, M.D., for editorial comments and Jill Garroway, Rebecca Mel- rose, Nathan Stein, Mari Sohma, and Shuna Klaviness for their assis- may reflect developmental or genetic influences or envi- tance with the study. ronmental insults (e.g., hypoxia) that have led to reductions in what were previously normally developing volumes. This hypothesis suggests that two hits (genetic and References environmental), occurring either independently or in in- 1. Wozniak J, Biederman J, Kiely K, Ablon JS, Faraone SV, Mundy teraction, may result in reduction of hippocampal vol- E, Mennin D: Mania-like symptoms suggestive of childhood-on- umes as well as in the expression of bipolar disorder. set bipolar disorder in clinically referred children. J Am Acad This report describes to our knowledge the largest MRI Child Adolesc Psychiatry 1995; 34:867–876 study of pediatric bipolar disorder to date. However, the 2. Geller B, Luby J: Child and adolescent bipolar disorder: a review of the past 10 years. J Am Acad Child Adolesc Psychiatry number of subjects was still relatively small and repre- 1997; 36:1168–1176 sents only a cross-sectional look at youths with bipolar 3. Geller B, Craney JL, Bolhofner K, DelBello MP, Williams M, disorder and comparison subjects. Our findings should be Zimerman B: One-year recovery and relapse rates of children considered in light of other limitations, such as the smaller with a prepubertal and early adolescent bipolar disorder phe- number of comparison subjects relative to the number of notype. Am J Psychiatry 2001; 158:303–305 4. Lewinsohn PM, Klein DN, Seeley JR: Bipolar disorders in a com- youths with bipolar disorder, the lack of rating scales for munity sample of older adolescents: prevalence, phenomenol- depressive symptoms, and the difficulty in reliably deter- ogy, comorbidity, and course. J Am Acad Child Adolesc Psychi- mining the age at onset of a child or adolescent’s mood in- atry 1995; 34:454–463 stability based on parental recall. Although we did exam- 5. Johnson L, Andersson-Lundman G, Åberg-Wistedt A, Mathé AA: ine the effects of several clinical variables, including age Age of onset in affective disorder: its correlation with heredi- and type of medication as well as chlorpromazine equiva- tary and psychosocial factors. J Affect Disord 2000; 59:139–148 6. Goldstein JM, Seidman LJ, O’Brien LM, Horton NJ, Kennedy DN, lents, and found no significant effects, the power in our Makris N, Caviness VS Jr, Faraone SV, Tsuang MT: Impact of nor- study may have been insufficient to fully assess the possi- mal sexual dimorphisms on sex differences in structural brain ble effects of these parameters. abnormalities in schizophrenia assessed by magnetic reso- Our findings support the hypothesis that the hippocam- nance imaging. Arch Gen Psychiatry 2002; 59:154–164 7. Reiter EO, Fuldauer VG, Root AW: Secretion of the adrenal an- pus may be involved in the underlying pathophysiology of drogen, dehydroepiandrosterone sulfate, during normal in- pediatric bipolar disorder and may represent a unique fancy, childhood, and adolescence, in sick infants, and in chil- characteristic of early-onset presentation of the disorder dren with endocrinologic abnormalities. J Pediatr 1977; 90: due to a genetic diathesis or derangements in growth pro- 766–770 cesses around the time of illness onset. However, in order 8. Soares JC, Mann JJ: The anatomy of mood disorders: review of to obtain a better sense of the trajectory of abnormalities structural neuroimaging studies. Biol Psychiatry 1997; 41:86– 106 in this structure in bipolar disorder populations, longitudi- 9. Phillips ML, Drevets WC, Rauch SL, Lane R: The neurobiology of nal studies are needed. Future studies need large enough emotion perception, I: the neural basis of normal emotion sample sizes of both sexes and a sex-matched healthy perception. Biol Psychiatry 2003; 54:504–514

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1263 NEUROIMAGING FINDINGS IN PEDIATRIC BIPOLAR DISORDER

10. Phillips ML, Drevets WC, Rauch SL, Lane RD: The neurobiology 30. Caviness VS Jr, Meyer J, Makris N, Kennedy DN: MRI-based topo- of emotion perception, II: implications for major psychiatric graphic parcellation of human neocortex: an anatomically disorders. Biol Psychiatry 2003; 54:515–528 specified method with estimate reliability. J Cogn Neurosci 11. Packard MG, Cahill L: Affective modulation of multiple mem- 1996; 8:566–587 ory systems. Curr Opin Neurobiol 2001; 11:752–756 31. Kennedy DN, Filipek PA, Caviness VS Jr: Autonomic segmenta- 12. Altshuler LL, Bartzokis G, Grieder T, Curran J, Jimenez T, Leight tion and volumetric calculations in nuclear magnetic reso- K, Wilkins J, Gerner R, Mintz J: An MRI study of temporal lobe nance imaging. IEEE Trans Med Imaging 1989; 8:1–7 structures in men with bipolar disorder or schizophrenia. Biol 32. Worth AJ, Makris N, Caviness VS Jr, Kennedy DN: Neuroanatom- Psychiatry 2000; 48:147–162 ical segmentation in MRI: technological objectives. Intern J Pat- 13. Strakowski SM, DelBello MP, Sax KW, Zimmerman ME, Shear tern Recognit Artif Intell 1997; 11:1161–1187 PK, Hawkins JM, Larson ER: Brain magnetic resonance imaging 33. Filipek PA, Richelme C, Kennedy DN, Caviness VS Jr: The young of structural abnormalities in bipolar disorder. Arch Gen Psy- adult human brain: an MRI-based morphometric study. Cereb chiatry 1999; 56:254–260 Cortex 1994; 4:344–360 14. Pearlson GD, Barta PE, Powers RE, Menon RR, Richards SS, Ayl- 34. Seidman LJ, Faraone SV, Goldstein JM, Goodman JM, Kremen ward EH, Federman EB, Chase GA, Petty RG, Tien AY: Medial WS, Matsuda G, Hoge EA, Kennedy D, Makris N, Caviness VS Jr, and superior temporal gyral volumes and cerebral asymmetry Tsuang MT: Reduced subcortical brain volumes in nonpsy- in schizophrenia versus bipolar disorder. Biol Psychiatry 1997; chotic siblings of schizophrenic patients: a pilot magnetic reso- 41:1–14 nance imaging study. Am J Med Genet 1997; 74:507–514 15. Dupont RM, Jernigan TL, Heindel W, Butters N, Shafer K, Wilson 35. Seidman LJ, Faraone SV, Goldstein JM, Goodman JM, Kreman T, Hesselink J, Gillin JC: Magnetic resonance imaging and mood WS, Toomey R, Tourville J, Kennedy D, Makris N, Caviness VS, disorders: localization of white matter and other subcortical Tsuang MT: Thalamic and amygdala-hippocampal volume re- abnormalities. Arch Gen Psychiatry 1995; 52:747–755 ductions in first-degree relatives of patients with schizophre- 16. Swayze VW II, Andreasen NC, Alliger RJ, Yuh WT, Ehrhardt JC: nia: an MRI-based morphometric analysis. Biol Psychiatry Subcortical and temporal structures in affective disorder and 1999; 46:941–954 schizophrenia: a magnetic resonance imaging study. Biol Psy- 36. Seidman LJ, Faraone SV, Goldstein JM, Kreman WS, Horton NH, chiatry 1992; 31:221–240 Makris N, Toomey R, Kennedy D, Caviness VS, Tsuang MT: Left 17. Noga JT, Vladar K, Torrey EF: A volumetric magnetic resonance hippocampal volume as a vulnerability indicator for schizo- imaging study of monozygotic twins discordant for bipolar dis- phrenia. Arch Gen Psychiatry 2002; 59:839–849 order. Psychiatry Res 2001; 106:25–34 37. Kates WR, Abrams MT, Kaufmann WE, Breiter SN, Reiss AL: Re- 18. Blumberg HP, Kaufman J, Martin A, Whiteman R, Zhang JH, liability and validity of MRI measurement of the amygdala and Gore JC, Charney DS, Krystal JH, Peterson BS: Amygdala and hippocampus in children with fragile X syndrome. Psychiatry hippocampal volumes in adolescents and adults with bipolar Res 1997; 75:31–48 disorder. Arch Gen Psychiatry 2003; 60:1201–1208 38. Caviness VS Jr, Kennedy DN, Richelme C, Rademacher J, Filipek 19. Botteron KN, Vannier MW, Geller B, Todd RD, Lee BC: Prelimi- PA: The human brain age 7–11 years: a volumetric analysis nary study of magnetic resonance imaging characteristics in 8- based on magnetic resonance images. Cereb Cortex 1996; 6: to 16-year-olds with mania. J Am Acad Child Adolesc Psychiatry 726–736 1995; 34:742–749 39. Lange N, Giedd JN, Castellanos FX, Vaituzis AC, Rapoport JL: 20. Friedman L, Findling RL, Kenny JT, Swales TP, Stuve TA, Jes- Variability of human brain structure size: ages 4–20 years. Psy- berger JA, Lewin JS, Schulz SC: An MRI study of adolescent pa- chiatry Res 1997; 74:1–12 tients with either schizophrenia or bipolar disorder as com- 40. Durston S, Hulshoff Pol HE, Casey BJ, Giedd JN, Buitelaar JK, van pared to healthy control subjects. Biol Psychiatry 1999; 46:78– Engeland H: Anatomical MRI of the developing human brain: 88 what have we learned? J Am Acad Child Adolesc Psychiatry 21. Dasari M, Friedman L, Jesberger J, Stuve TA, Findling RL, Swales 2001; 40:1012–1020 TP, Schulz SC: A magnetic resonance imaging study of thalamic 41. Hauser P, Matochik J, Altshuler LL, Denicoff KD, Conrad A, Li X, area in adolescent patients with either schizophrenia or bipo- Post RM: MRI-based measurements of temporal lobe and ven- lar disorder as compared to healthy controls. Psychiatry Res tricular structures in patients with bipolar I and bipolar II disor- 1999; 91:155–162 ders. J Affect Disord 2000; 60:25–32 22. DelBello MP, Zimmerman ME, Mills NP, Getz GE, Strakowski SM: 42. Benes FM, Kwok EW, Vincent SL, Todtenkopf MS: A reduction of Magnetic resonance imaging analysis of amygdala and other nonpyramidal cells in sector CA2 of schizophrenics and manic subcortical brain regions in adolescents with bipolar disorder. depressives. Biol Psychiatry 1998; 44:88–97 Bipolar Disord 2004; 6:43–52 43. Fatemi SH, Earle JA, Stary JM, Lee S, Sedgewick J: Altered levels 23. Orvaschel H, Puig-Antich J: Schedule for Affective Disorders of the synaptosomal associated protein SNAP-25 in hippocam- and Schizophrenia for School-Age Children—Epidemiologic pus of subjects with mood disorders and schizophrenia. Neu- Version (K-SADS-E), 4th revision. Fort Lauderdale, Fla, Nova roreport 2001; 12:3257–3262 University, Center for Psychological Studies, 1987 44. Fatemi SH, Earle JA, McMenomy T: Reduction in Reelin immu- 24. Tanner JM: Growth at Adolescence. Oxford, UK, Blackwell Sci- noreactivity in hippocampus of subjects with schizophrenia, entific Publications, 1962 bipolar disorder and major depression. Mol Psychiatry 2000; 5: 25. Wechsler D: Manual for the Wechsler Intelligence Scale for Chil- 654–663 dren. San Antonio, Tex, Psychological Corp (Harcourt), 1991 45. Eastwood SL, Harrison PJ: Hippocampal synaptic pathology in 26. Oldfield RC: The assessment and analysis of handedness: the schizophrenia, bipolar disorder and major depression: a study Edinburgh Inventory. Neuropsychologia 1971; 9:97–113 of complexin mRNAs. Mol Psychiatry 2000; 5:425–432 27. Young RC, Biggs JT, Ziegler VE, Meyer DA: A rating scale for ma- 46. Dowlatshahi D, MacQueen G, Wang JF, Chen B, Young LT: In- nia: reliability, validity and sensitivity. Br J Psychiatry 1978; creased hippocampal supragranular Timm staining in subjects 133:429–435 with bipolar disorder. Neuroreport 2000; 11:3775–3778 28. Woods SW: Chlorpromazine equivalent doses for the newer 47. Law AJ, Deakin JF: Asymmetrical reductions of hippocampal atypical antipsychotics. J Clin Psychiatry 2003; 64:663–667 NMDAR1 glutamate receptor mRNA in the psychoses. Neurore- 29. Stoll AL: The Psychopharmacology Reference Card, 2001 port 2001; 12:2971–2974

1264 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FRAZIER, CHIU, BREEZE, ET AL.

48. Rosoklija G, Toomayan G, Ellis SP, Keilp J, Mann JJ, Latov N, Hays 54. Hellgren L, Gillberg IC, Bagenholm A, Gillberg C: Children with AP, Dwork AJ: Structural abnormalities of subicular dendrites in deficits in attention, motor control and perception (DAMP) al- subjects with schizophrenia and mood disorders: preliminary most grown up: psychiatric and personality disorders at age 16 findings. Arch Gen Psychiatry 2000; 57:349–356 years. J Child Psychol Psychiatry 1993; 35:1255–1271 49. Goldstein JM, Seidman LJ, Horton NJ, Makris N, Kennedy D, Cav- 55. Frazier JA, Biederman J, Tohen M, Feldman PD, Jacobs TG, iness VS, Faraone S, Tsuang MT: Normal sexual dimorphism of Toma V, Rater MA, Tarazi RA, Kim GS, Garfield SB, Sohma M, the adult human brain assessed by in vivo magnetic resonance Gonzalez-Heydrich J, Risser RC, Nowling ZM: A prospective imaging. Cereb Cortex 2001; 11:490–497 50. DelBello MP, Zimerman ME, Getz GE, Strakowski S: Structural open-label treatment trial of olanzapine monotherapy in chil- MRI analysis of adolescents with bipolar disorder. Biol Psychia- dren and adolescents with bipolar disorder. J Child Adolesc Psy- try 2001; 49:20S–21S chopharmacol 2001; 11:239–250 51. Kennedy DN, Lange N, Makris N, Bates J, Meyer J, Caviness VS 56. Tsuang M: Schizophrenia: genes and environment. Biol Psychi- Jr: Gyri of the human neocortex: an MRI-based analysis of vol- atry 2000; 47:210–220 ume and variance. Cereb Cortex 1998; 8:372–384 57. Todd RD, Reich W, Petti TA, Joshi P, DePaulo JR Jr, Nurnberger J 52. Pantelis C, Velakoulis D, McGorry PD, Wood SJ, Suckling J, Phil- Jr, Reich T: Psychiatric diagnoses in the child and adolescent lips LJ, Yung AR, Bullmore ET, Brewer W, Soulsby B, Desmond P, members of extended families identified through adult bipolar McGuire PK: Neuroanatomical abnormalities before and after affective disorder probands. J Am Acad Child Adolesc Psychia- onset of psychosis: a cross-sectional and longitudinal MRI comparison. Lancet 2003; 361:281–288 try 1996; 35:664–671 53. Sigurdsson E, Fombonne E, Sayal K, Checkley S: Neurodevelop- 58. Goldstein JM, Kennedy DN, Caviness VS Jr: Brain development, mental antecedents of early-onset bipolar affective disorder. XI: sexual dimorphism (image, neuro). Am J Psychiatry 1999; Br Med J 1999; 174:121–127 156:352

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1265 Article

Familial Variation in Episode Frequency in Bipolar Affective Disorder

Maria E. Fisfalen, M.D. Objective: Bipolar affective disorder is a isodes/year. Episode frequency was signifi- familial illness characterized by recurrent cantly correlated among relatives (r=0.56, Thomas G. Schulze, M.D. episodes of mania and depression, but lit- p<0.004). Earlier age at onset, bipolar II tle is known about the familial nature of disorder, hallucinations or delusions, alco- J. Raymond DePaulo, Jr., M.D. episode recurrence or its associated clini- holism, and suicidal behavior were all cal features. The authors analyzed the re- more prevalent in the highest than in the currence frequency of affective episodes lowest quartiles of episode frequency. Fe- Leslie J. DeGroot, M.D. (episode frequency), along with associ- male gender and recurrent major depres- ated clinical and demographic variables, sion were more prevalent in the lowest Judith A. Badner, M.D. in families with at least three members quartile. Panic disorder, substance abuse, with a major affective disorder. and thyroid disease were all unrelated to Francis J. McMahon, M.D. Method: Members of 86 families ascer- episode frequency. Subjects with DSM-IV tained through probands with bipolar af- rapid cycling did not differ from other affective disorder who had two or more fected subjects for most of the variables first-degree relatives with a major affec- tested. tive disorder were interviewed by psychia- Conclusions: Episode frequency is a trists and assigned an all-sources diagno- highly familial trait in bipolar affective sis. Data for 407 subjects with a major disorder, associated with several indica- affective disorder were analyzed. Episode tors of severity, and may be useful in de- frequency was estimated as the number fining clinical subtypes of bipolar affective of episodes of major depression, mania, disorder with greater genetic liability. and hypomania per year of illness. DSM-IV rapid cycling was not supported Results: Episode frequency was smoothly by these data as the best predictor of fa- distributed over the range of 0.02–20.2 ep- miliality or severity.

(Am J Psychiatry 2005; 162:1266–1272)

A ll patients with bipolar affective disorder experience quency. This issue is important, because the absence of fa- recurrent episodes of major depression, mania, hypoma- miliality could suggest that episode frequency is under the nia, or mixed states, but the frequency with which epi- primary control of nongenetic factors. We report here an sodes recur can range from one in several years to many analysis of episode frequency in families ascertained for a per day (1). As many as 20% of people with bipolar affec- genetic linkage study of bipolar affective disorder. We also tive disorder experience rapid cycling, defined by DSM-IV examined the relationship between episode frequency and as four or more affective episodes in a year (1, 2). Rapid cy- age at onset, suicidal behavior, psychosis, panic disorder, cling is more common in women and in people with bipo- alcoholism, substance abuse, and thyroid disease in these families. We found that episode frequency is a highly famil- lar II disorder (1, 3–5). Hypothyroidism, steroid hormones, ial trait, associated with several indicators of severity, that and the use of antidepressants have been associated with may help define clinical subtypes of bipolar affective disor- rapid cycling, but these findings are controversial (6–10). der with the greatest genetic liability. Rapid cycling is just one extreme of the spectrum of episode frequency. Few studies have examined the full range of episode frequency in bipolar affective disorder. Method Similarly, although bipolar affective disorder is a familial Subjects illness, the nature of episode frequency as a familial trait A total of 625 subjects in 86 families ascertained for a genetic has not been investigated extensively (1, 11). Several stud- linkage study (16) were eligible for the current analysis. All study ies have addressed morbid risk for bipolar disorder among volunteers gave written informed consent after the procedures the relatives of rapid-cycling probands (4, 5, 12, 13), and had been fully explained. some studies have addressed the tendency of rapid cycling Families were ascertained through a proband with a reported history of bipolar I disorder and at least two first-degree relatives to run in families (14, 15), but we are aware of no previous (at least two siblings or at least one sibling and only one parent) studies that directly measure the familiality of episode fre- with a major affective disorder.

1266 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FISFALEN, SCHULZE, DEPAULO, ET AL.

TABLE 1. Characteristics of Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder (N=86) and Their Relatives With a Major Affective Disorder (N=321) in a Study of Familial Variation in Episode Frequency, by Diagnostic Group Age at Duration Number of Number of Number of Number of Age at Onset Interview of Illness Episodes of Episodes of Episodes of Episodes Gender (years) (years) (years) Hypomania Mania Depression per Yeara Diagnostic group M F MeanSDMeanSDMeanSDMeanSDMeanSDMeanSDMeanSD Bipolar I disorder (N=144) 59 85 21.1 9.6 41.7 13.3 20.6 10.9 27.8 36.2 7.5 11.7 10.1 14.3 1.0 1.1 Bipolar II disorder (N=151) 57 94 20.4 9.1 44.2 15.5 23.7 12.6 47.9 41.1 — — 17.3 28.0 3.1 3.0 Recurrent unipolar depression (N=102)b 23 79 25.9 12.1 43.4 14.9 17.5 11.8 — — — — 5.7 11.9 0.4 0.5 Schizoaffective disorder, bipolar type (N=10) 7 3 22.4 12.2 40.9 12.9 18.5 10.4 — — 5.6 6.9 9.3 11.4 0.7 0.8 All diagnoses (N=407) 146 261 22.1 10.4 43.1 14.5 20.9 12.0 45.9 40.0 7.4 11.8 11.7 11.4 1.6 2.3 a Significant overall difference among diagnostic groups (p≤0.0001, analysis of variance). In paired comparisons, significant differences between bipolar II disorder and bipolar I disorder groups (t=7.8, df=192, p<0.0001), bipolar II disorder and recurrent unipolar depression groups (t=10.5, df=165, p<0.0001), bipolar II disorder and schizoaffective disorder, bipolar type, groups (t=6.4, df=30, p<0.0001), and bipolar I disorder and recurrent unipolar depression groups (t=5.5, df=222, p<0.0001). b Significant difference between the proportion of male and female subjects (χ2=10.5, df=1, p=0.001).

FIGURE 1. Number of Depressive and Manic Episodes per Year Since Onset of Major Affective Disorder in Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder (N=86) and Their Relatives With Major Affective Disorder (N=321)a

80 Depressive 70 Manic 60 50 40 30 20 Percent of Subjects Percent 10 0 0.0 0.6 1.2 1.8 2.4 3.0 3.6 4.2 4.8 5.4 6.0 6.5 >6.5

Episodes per Year a The distribution of hypomanic episodes per year of illness was similar (data not shown).

Among affected subjects, the average level of education was 27.8% (N=174) were deemed unaffected. Of the 407 affected sub- 14.3 years (SD=3.0), and 77.3% were employed. Twenty-six per- jects, 64% were female. cent were married, 55.8% were separated or divorced, and the rest Retrospective lifetime self-report data for several clinical vari- were widowed or never married. ables were collected with the SADS-L. This study focused on age at onset and number of episodes of mania, hypomania, and ma- Clinical Assessment jor depression, which were derived directly from the correspond- All subjects were interviewed by a psychiatrist who used the ing items in the SADS-L instrument. Age at onset was defined as age at first mania or major depression, whichever was earlier (20). Schedule for Affective Disorders and Schizophrenia—Lifetime The number of episodes was estimated by the interviewer, who Version (SADS-L) (17). Two additional psychiatrists reviewed the worked with each subject to define in detail the most severe epi- interview notes, family informant data, and medical records be- sode of mania, hypomania, and major depression, then asked, fore assigning best-estimate diagnoses under Research Diagnos- “How many episodes like this have you had in your lifetime?” tic Criteria (18). The diagnosis of bipolar II disorder required re- Subjects were asked to count periods of illness separated by at current major depression as well as hypomania. All diagnoses least 2 months of recovery as separate episodes. Continuous peri- were found to be highly reliable, on the basis of assessments of ods of illness involving a switch in polarity were counted as two co-rated and test-retest interviews, as well as agreement between episodes, provided that each pole appeared to fulfill the diagnos- the psychiatrists who provided the best-estimate diagnoses (19). tic criteria, but these switches were uncommon. Final diagnoses among the probands were as follows: bipolar I Data were also extracted for six clinical variables that have disorder, N=71; bipolar II disorder, N=12; and schizoaffective dis- been associated with bipolar affective disorder in the literature: order, bipolar type, N=3. In the total study group of probands and alcoholism, substance abuse, panic disorder, psychosis, suicidal their family members, 23.0% (N=144) had bipolar I disorder, behavior, and thyroid disease (1). Thyroid disease data (hypo- or 24.2% (N=151) had bipolar II disorder, 16.2% (N=102) had recur- hyperthyroidism) were not collected during the first year of the rent unipolar depression, 1.6% (N=10) had schizoaffective disor- study but were available from 312 affected and 149 unaffected der, bipolar type, 7.0% (N=44) had an uncertain diagnosis, and subjects ascertained later.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1267 EPISODE FREQUENCY IN BIPOLAR DISORDER

TABLE 2. Clinical Features of Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder and Their Relatives With Major Affective Disorder: Comparison Between Lowest and Highest Quartiles of Episode Frequency Subjects in the Lowest Quartile Subjects in the Highest Quartile Clinical Feature (N=98) (N=101) Analysis Mean SD Range Mean SD Range t df p

Episodes/year 0.16 0.06 0.02–0.28 4.69 2.88 2.3–20.2 15.7 197 <0.0001 Age at onset (years) 27.9 11.6 18.2 7.0 7.0 197 <0.0001

N % N % χ2 df p Diagnosisa Bipolar I disorder 26 26.5 20 19.8 n.s. Bipolar II disorder 8 8.2 78 77.2 96.6 1 ≤0.0001 Recurrent unipolar depression 60 61.2 2 2.0 81.3 1 ≤0.0001 Schizoaffective disorder, bipolar type 4 4.1 1 1.0 n.s. Gender 4.9 1 <0.04 Female 71 72.4 58 57.4 Male 27 27.6 43 42.6 Axis I comorbidity Panic disorder 10 10.2 17 16.8 n.s. Alcoholism 21 21.4 43 42.6 10.1 1 0.002 Substance abuse/dependence 11 11.2 18 17.8 n.s. Suicidal behavior 6 6.1 21 20.8 9.1 1 0.003 a Significant overall difference between patients in the lowest and highest quartiles (χ2=113.7, df=1, p≤0.0001).

Statistical Analysis Episode frequency was smoothly distributed over the Episode frequency was calculated on the basis of the total re- range of 0.02 to 20.2 episodes per year in this group of sub- ported episodes of major depression, mania, or hypomania per jects (Figure 1). We observed no discontinuity at the DSM- year of illness, as follows: episode frequency = total number of ep- IV rapid-cycling threshold of four episodes per year. The isodes of illness/(age at interview – age at onset). Five subjects distributions were similar when depressive, manic, and with an illness duration of less than 2 years were excluded, because a small denominator may inflate the episode frequency and hypomanic episodes were considered separately, so only subjects may tend to volunteer for a study around a time of in- the total episode frequency was considered in the subse- creased illness activity. Reports of mixed episodes are not elicited quent analyses. by the SADS-L, so data for mixed episodes are not included in this calculation. Because episode frequency showed a highly skewed Association With Diagnostic Group distribution, the values were log-transformed. The data were stored by using a relational database system that Episode frequency was significantly associated with di- was based on Paradox (versions 5 and 8) (Corel Corp., Ottawa, agnostic group (p<0.001). Specific comparisons revealed Canada) (21). that episode frequency was highest among subjects with Continuous variables were analyzed by t test or analysis of vari- bipolar II disorder and lowest among subjects with recur- ance (ANOVA), and categorical variables were analyzed by Pear- rent unipolar depression (Bonferroni corrected p<0.001 son’s chi-square test and Fisher’s exact test. Because the use of relatives’ data in the ANOVA could lead to biased estimates of for each of four comparisons). When subjects in the lowest variance, familiality was also assessed in a mixed-effects regres- and highest quartiles of episode frequency (range=0.02– sion procedure in which the likelihood ratio test was used to com- 0.28 episodes/year and range=2.3–20.2 episodes/year, re- pare the log likelihoods of “intercept-only” models and models spectively) were contrasted (Table 2), recurrent unipolar that include family membership as a random effect (22). The depression was more prevalent in the lowest quartile, and mixed-effects model assumes that data within clusters are dependent (as is the case for data from relatives) and estimates the bipolar II disorder was more prevalent in the highest quar- degree of dependency along with the parameters of the model. tile (p≤0.0001). The ANOVA was performed with SAS (SAS Institute, Cary, N.C.), and the regression analyses were performed with MIXREG (23). Association With Other Variables The alpha level was set at 0.05. All tests were two-tailed. Gender. There was no significant difference in episode frequency between male and female subjects (t=1.39, df= Results 403, n.s.), but subjects in the lowest quartile of episode fre- Descriptive Statistics quency were more likely to be female (Table 2). The variables used in the analysis of episode frequency, Age at onset. Age at onset was negatively correlated along with their relationship to gender and diagnostic with episode frequency (r=–0.19, N=401, p<0.001). How- group, are shown in Table 1. There were no significant dif- ever, the relationship was not strictly linear (Figure 2). The ferences among diagnostic groups in mean age at onset, highest episode frequencies were seen among subjects duration of illness, or gender, except for the expected with onset between ages 15 and 18 years. Similarly, age at higher prevalence of women in the recurrent unipolar de- onset was significantly lower among those in the highest pression group. quartile of episode frequency (Table 2).

1268 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FISFALEN, SCHULZE, DEPAULO, ET AL.

Psychosis. Psychosis, defined here as hallucinations or FIGURE 2. Relationship of Age at Onset of Major Affective delusions (1), frequently complicated bipolar affective Disorder With Episode Frequency in Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder (N=86) and Their disorder in this group of subjects; 31.2% of subjects re- Relatives With Major Affective Disorder (N=321) ported a history of psychosis at some time during their ill- 24 ness. Episode frequency was significantly higher in subjects with a history of psychosis than in those with no such history (F=5.53, df=1, 392, p<0.02). 20 Suicidal behavior and other axis I comorbidities. Typical patterns of suicidal behavior and comorbidity were observed in this group of subjects (1, 24, 25). Alcoholism was the most common comorbid disorder, found in 33.8% 16 (N=136) of the subjects. Suicidal behavior was present in 20.6% (N=83), illicit substance abuse/dependence in 16.7% (N=67), and panic disorder in 11.9% (N=48). 12 Episode frequency was significantly associated with some but not all of these variables. Episode frequency was Episodes per Year significantly associated with alcoholism (F=8.64, df=1, 8 400, p=0.003) and with a history of suicidal behavior (F= 7.23, df=1, 400, p=0.007). Panic disorder and illicit substance abuse/dependence were not associated with epi- 4 sode frequency (F=2.49, df=1, 397, n.s., and F=2.35, df=1, 400, n.s., respectively). Similar results were observed in the quartiles analysis (Table 2), which also demonstrates the direction of each association. 0 020406080 Familiality Age at Onset (years) Episode frequency was significantly correlated among probands and their affected relatives (intraclass r=0.56, F= our findings. The proportion of female subjects was simi- 1.53, df=96, 321, p<0.004), suggesting that more than 30% lar in the groups with and without thyroid disease, and the of the variance in episode frequency was accounted for by mean age at interview was lower among affected subjects family membership. with thyroid disease than among unaffected subjects The familiality of episode frequency was confirmed in (mean=46.6 years, SD=11.9, versus mean=68.5 years, SD= < the mixed regression analysis (Table 3), which accounted 11.0) (t=3.9, df=30, p 0.001). for the nonindependence of data among relatives. Episode DSM-IV rapid cycling frequency remained strongly familial when the correlated variables (diagnostic group, age at interview, age at onset, Forty-six subjects (11.1%) met the DSM-IV criteria for psychosis, alcoholism, and suicidal behavior) were in- rapid cycling (four or more episodes/year). These subjects cluded in the model as fixed effects. This finding showed were significantly more likely than those reporting fewer that the familiality of episode frequency in this group of than four episodes/year to have a diagnosis of bipolar II subjects was not due solely to any familial tendency of the disorder (87.1%) (χ2=55.2, df=1, p<0.0001). However, none correlated variables. of the other demographic or clinical variables analyzed in this study differed between the subjects with DSM-IV Thyroid disease comorbidity rapid cycling and the other subjects. We did not detect sig- There was no significant difference in episode fre- nificant evidence that the categorical trait of rapid cycling quency between subjects with thyroid disorder and those was familial in this study group. without thyroid disorder. However, 43 (13.8%) of 312 affected subjects and nine (6%) of 149 unaffected subjects Discussion reported a history of thyroid disease (Table 4). A similar proportion of affected subjects who were exposed to lith- To our knowledge, this study demonstrated for the first ium (19 [14.3%] of 132) and affected subjects who were not time that episode frequency, defined as the number of exposed to lithium (24 [13.3%] of 180) reported thyroid affective episodes per year, is a familial trait in bipolar af- disease, and both of those groups were significantly more fective disorder. We further showed that the familiality of likely to report thyroid disease than were the unaffected episode frequency is not accounted for by other, corre- subjects (p<0.05). The prevalence of thyroid disease is lated variables, such as affective diagnosis subtype, psy- generally considered to be greater in females and to in- chosis, alcoholism, or suicidal behavior. As a familial trait crease with age (26), but this pattern did not account for associated with several indicators of disease severity, epi-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1269 EPISODE FREQUENCY IN BIPOLAR DISORDER

TABLE 3. Familiality of Episode Frequency in the Families of 86 Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder

Log-Likelihood Log-Likelihood Likelihood Ratio Statistic Intraclass Correlation Fixed Effects Random Effect: Family No Random Effect Ratioa (χ2, df=1) p Coefficientb None –326.34 –338.97 1.04 25.28 <0.0001 0.12 Correlatesc –269.16 –299.63 1.11 60.93 <0.0001 0.13 a Values >1 favor the model with family as a random effect. b A measure of the strength of association within families. c Includes diagnostic group, age at interview, age at onset, suicidal behavior, alcoholism, and psychosis.

TABLE 4. Thyroid Disease in Families of 86 Probands With Bipolar I, II, and Schizoaffective Bipolar Disorder in a Study of Familial Variation in Episode Frequency Affected Subjects Affected Subjects Not Taking Lithium Unaffected Subjects (N=312) (N=180) (N=149) Thyroid Disease Group N % N % N % Subjects with no thyroid disease 269 86.2 156 86.6 140 94.0 Subjects with hypo/hyperthyroidisma 43a 13.8 24b 13.3 9 6.0 a Significant difference in the proportion of subjects with thyroid disease between affected subjects and unaffected subjects (χ2=6.0, df=1, p<0.02) and between affected subjects not taking lithium and unaffected subjects (χ2=4.8, df=1, p<0.05). sode frequency may help to define clinical subtypes of bi- rapid cycling punctuating a course of illness with few polar affective disorder with greater genetic liability. These other episodes. We cannot rule out familial effects in these data did not support DSM-IV rapid cycling as the best pre- subjects. dictor of familiality or severity. The evidence linking hypothyroidism and rapid cycling The data were collected from 407 subjects in 86 families, is controversial. The association of hypothyroidism with to our knowledge the largest data set ever studied for epi- major affective disorders and the presence of thyroid anti- sode frequency. Diagnoses were highly reliable and were bodies in patients with major affective disorders have made by including all available data in a best-estimate been described in numerous publications, but the biolog- procedure. The primary data used in estimating episode ical relationship between these two illnesses remains un- frequency were retrospective and thus subject to recall clear (6–8, 10). Our data, which relied solely on subjects’ bias. There was no reason to expect, however, that rela- self-report, were limited by lack of information from thy- tives would be correlated in their recall bias; thus, recall roid function tests and thyroid antibodies assays. Never- bias alone cannot account for our findings. The data col- theless, the rate of reported thyroid disease among unaf- lection methods had other limitations. Subjects with two fected subjects (6%) appeared similar to the population or more episodes of mania were not questioned in detail rate of 5.9% (26), and the rate of reported thyroid disease about all hypomanic episodes. Thus, true episode fre- among the lithium-treated subjects was similar to that re- quency was probably underestimated for subjects with bi- ported in another study (27). Thus, we do not appear to polar I disorder. This underestimation was also unlikely to have greatly under- or overestimated the rates of thyroid account for our main findings, because the exclusion of disease in these study subjects. We found, as have others, hypomanic episodes did not change the distribution of that thyroid disease was more common among affected episode frequency in this study group. No data were col- subjects and that this difference remained significant even lected on treatment, so it was not possible to control for after subjects with a history of lithium exposure were potential treatment effects on episode frequency in these data. dropped from the analysis. We found no association between thyroid disease and episode frequency in the study DSM-IV defines subjects with four or more major affec- group, consistent with previous studies (4, 5, 28). tive episodes in a year as having rapid cycling. The published studies of rapid-cycling subjects, so defined, have Alcoholism and substance abuse are known to be highly not consistently detected familial aggregation (4, 5, 12– associated with major affective disorders (1, 24, 25). In a 15). We found little support for the DSM-IV definition of previous study that included some of the subjects also in- rapid cycling in this study group. As a categorical trait, cluded in the present analysis, alcoholism was found to be rapid cycling was not familial in this analysis, even though associated with a higher rate of suicide attempts and to be we found significant evidence of familiality when we con- clustered in a subset of families (29). The present study ex- sidered subjects across the full range of episode frequency. tended this finding by demonstrating that alcoholism and Rapid cycling was associated with the diagnosis of bipolar suicide attempts are both strongly associated with episode II disorder, but it was not correlated with any of the other frequency. Our new results suggest that the association clinical features we examined. Our retrospective data did between alcoholism and suicidal behavior is mediated, at not allow us to identify subjects with discrete periods of least in part, by an increase in episode frequency.

1270 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 FISFALEN, SCHULZE, DEPAULO, ET AL.

Episode frequency was also associated with other clinical features of bipolar affective disorder. Age at onset (it- Presented in part at the 57th annual meeting of the Society of Bio- logical Psychiatry, Philadelphia, May 16–18, 2002. Received Jan. 29, self a significant predictor of prognosis, comorbidity, and 2003; revision received June 24, 2004; accepted Aug. 2, 2004. From treatment response in bipolar affective disorder [30]) was the Departments of Psychiatry and Medicine, University of Chicago, strongly associated with episode frequency in this study Chicago; the Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine, Baltimore; Division of group. We also found that episode frequency was signifi- Genetic Epidemiology in Psychiatry, Central Institute of Mental cantly associated with psychotic features. It is possible Health, University of Heidelberg, Mannheim, Germany; and the Unit that episode frequency accounts for some of the tendency on the Genetic Basis of Mood and Anxiety Disorders, NIMH, Be- thesda, Md. Address correspondence and reprint requests to Dr. of psychotic features to run in families with bipolar affec- Fisfalen, Department of Psychiatry, Mount Sinai Medical Center, tive disorder (31, 32). An earlier analysis of a subset of Rosalind Franklin University of Medicine and Science, California at 15th St., Chicago, IL 60608; [email protected] (e-mail). these subjects indicated that episode frequency tended to Supported by grants from the NIMH Intramural Research Program increase in successive generations of a pedigree, a phe- and NIH, the National Alliance for Research on Schizophrenia and nomenon known as anticipation (33). A complete analysis Depression, the Chicago Brain Research Foundation, and the Edward F. Mallinckrodt, Jr., Foundation. Additional support for family recruit- of anticipation is beyond the scope of the study reported ment was provided by the Charles A. Dana Foundation. here, but if anticipation were present in the current study The authors thank Sylvia G. Simpson, Dean MacKinnon, and Melvin group, it would tend to decrease the familiality of episode G. McInnis for contributing to the family evaluations, Donald Hede- ker for advice on the use of MIXREG, and the family volunteers who frequency and thus could not account for our findings. make this work possible. It may come as a surprise that the highest quartile of episode frequency contained many subjects with bipolar II References disorder, which is traditionally considered less severe than bipolar I disorder, as well as many subjects with early on- 1. Goodwin FK, Jamison KR: Manic Depressive Illness. New York, set, psychotic features, alcoholism, and suicidal behavior, Oxford University Press, 1990 2. Dunner DL, Fieve RR: Clinical factors in lithium carbonate pro- clear indicators of a severe illness. This finding is consis- phylaxis failure. Arch Gen Psychiatry 1974; 30:229–233 tent with previous reports indicating more chronicity and a 3. Leibenluft E: Women with bipolar illness: clinical and research higher rate of comorbidity and suicidal behavior in bipolar issues. Am J Psychiatry 1996; 153:163–173 II disorder, compared to bipolar I disorder (34, 35). These 4. Coryell W, Endicott J, Keller M: Rapidly cycling affective disorder. Arch Gen Psychiatry 1992; 49:126–131 results imply that the traditional view of severity, which 5. Nurnberger J Jr, Guroff JJ, Hamovit J, Berrettini W, Gershon E: A emphasizes mania, may be too narrow. Bipolar II disorder family study of rapid-cycling bipolar illness. J Affect Disord is in many ways more “severe” than bipolar I disorder. 1988; 15:87–91 These findings have implications for genetic research in 6. Bauer MS, Whybrow PC: Rapid cycling bipolar affective disorder, I: association with grade I hypothyroidism. Arch Gen Psy- bipolar affective disorder. As a quantitative trait, episode chiatry 1990; 47:427–432 frequency may offer an alternative phenotype that would 7. Bartela L, Pellegrini L, Meschi M, Antonangeli L, Bogazzi F, be more powerful than the categorical phenotypes typi- Dell’Osso L, Pinchera A, Placidiet GF: Evaluation of thyroid cally used in genetic linkage and association studies. Epi- function in patients with rapid-cycling bipolar disorder. Psychi- sode frequency may also offer an approach to genetic het- atry Res 1990; 34:13–17 8. Haggerty JJ Jr, Evans DL, Golden RN, Pedersen CA, Simon JS, erogeneity in bipolar affective disorder, because subjects Nemeroff CB: The presence of antithyroid antibodies in pa- with similar episode frequencies may be more likely to tients with affective and nonaffective psychiatric disorders. share genetic determinants. Episode frequency is also as- Biol Psychiatry 1990; 27:51–60 sociated with several indicators of disease severity. To the 9. Esposito S, Prange AJ Jr, Golden RN: The thyroid axis and mood disorders: overview and future prospects. Psychopharmacol extent that disease severity is related to genetic liability, Bull 1997; 33:205–217 episode frequency may help to define clinical subtypes of 10. Kupka RW, Nolen WA, Post RM, McElroy SL, Altshuler LL, Deni- bipolar affective disorder with a greater burden of genetic coff KD, Frye MA, Keck PE Jr, Leverich GS, Rush AJ, Suppes T, Pol- risk factors. However, further data concerning the herita- lio C, Drexhage HA: High rate of autoimmune thyroiditis in bipolar disorder: lack of association with lithium exposure. Biol bility of episode frequency—for example, in twins—is Psychiatry 2002; 51:305–311 needed before we can make confident predictions about 11. Gershon ES, Hamovit J, Guroff JJ, Dibble E, Leckman JF, Sceery the potential value of episode frequency as a phenotype in W, Targum SD, Nurnberger JI Jr, Goldin LR, Bunney WE Jr: A genetic research. family study of schizoaffective, bipolar I, bipolar II, unipolar, and normal control probands. Arch Gen Psychiatry 1982; 39: In conclusion, we found that episode frequency is a fa- 1157–1167 milial trait in bipolar affective disorder. These data also 12. Maj M, Magliano L, Pirozzi R, Marasco C, Guarneri M: Validity of raise concerns about the DSM-IV definition of rapid cy- rapid cycling as a course specifier for bipolar disorder. Am J cling in bipolar affective disorder. We suggest that episode Psychiatry 1994; 151:1015–1019 13. Bauer MS, Calabrese J, Dunner DL, Post R, Whybrow PC, Gyulai frequency is an important clinical feature of bipolar affec- L, Tay LK, Younkin SR, Bynum D, Lavori P, Price RA: Multisite tive disorder, with implications for severity, comorbid data reanalysis of the validity of rapid cycling as a course mod- conditions, and genetic research. ifier for bipolar disorder. Am J Psychiatry 1994; 151:506–515

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1271 EPISODE FREQUENCY IN BIPOLAR DISORDER

14. MacKinnon DF, Zandi PP, Gershon E, Nurnberger JI Jr, Reich T, 26. Howell JG, Staehling NW, Flanders WD, Hannon WH, Gunter EW, DePaulo JR: Rapid switching of mood in families with multiple Spencer CA, Braverman LE: Serum TSH, T4 and thyroid anti- cases of bipolar disorder. Arch Gen Psychiatry 2003; 60:921– bodies in the United States population (1988 to 1994): Na- 928 tional Health and Nutrition Examination Survey (NHANES III). J 15. MacKinnon DF, Zandi PP, Gershon ES, Nurnberger JI Jr, DePaulo Clin Endocrinol Metab 2002; 87:489–499 JR Jr: Association of rapid mood switching with panic disorder 27. Johnston AM, Eagles JM: Lithium-associated clinical hypothy- and familial panic risk in familial bipolar disorder. Am J Psychi- roidism: prevalence and risk factors. Br J Psychiatry 1999; 175: atry 2003; 160:1696–1698 336–339 16. Simpson SG, Folstein SE, Meyers DA, DePaulo JR: Assessment of lineality in bipolar I linkage studies. Am J Psychiatry 1992; 149: 28. Post RM, Kramlinger KG, Joffe RT, Roy-Byrne PP, Rosoff A, Frye 1660–1665 MA, Huggins T: Rapid cycling bipolar affective disorder: lack of 17. Endicott J, Spitzer RL: A diagnostic interview: the Schedule for relation to hypothyroidism. Psychiatry Res 1997; 72:1–7 Affective Disorders and Schizophrenia. Arch Gen Psychiatry 29. Potash JB, Kane HS, Chiu Y-F, Simpson SG, MacKinnon DF, McIn- 1978; 35:837–844 nis MG, McMahon FJ, DePaulo JR Jr: Attempted suicide and al- 18. Spitzer RL, Endicott J, Robins E: Research Diagnostic Criteria: ra- coholism in bipolar disorder: clinical and familial relation- tionale and reliability. Arch Gen Psychiatry 1978; 35:773–782 ships. Am J Psychiatry 2000; 157:2048–2050 19. Simpson SG, McMahon FJ, McInnis MG, MacKinnon DF, Edwin 30. Coryell W, Solomon D, Turvey C, Keller M, Leon AC, Endicott J, D, Folstein SE, DePaulo JR: Diagnostic reliability of bipolar II dis- Schettler P, Judd L, Mueller T: The long-term course of rapid-cy- order. Arch Gen Psychiatry 2002; 59:736–740 cling bipolar disorder. Arch Gen Psychiatry 2003; 60:914–920 20. McMahon FJ, Stine OC, Chase GA, Meyers DA, Simpson SG, De- 31. Potash JB, Willour VL, Chiu Y-F, Simpson SG, MacKinnon DF, Paulo JR Jr: Influence of clinical subtype, sex, and lineality on Pearlson GD, DePaulo JR Jr, McInnis MG: The familial aggrega- age at onset of major affective disorder in a family sample. Am J Psychiatry 1994; 151:210–215 tion of psychotic symptoms in bipolar disorder pedigrees. Am 21. McMahon FJ, Thomas CJ, Koskela RJ, Breschel TS, Hightower T, J Psychiatry 2001; 158:1258–1264 Rohrer N, Savino C, McInnis MG, Simpson SG, DePaulo JR: Inte- 32. Potash JM, Chiu YF, Mackinnon DF, Miller EB, Simpson SG, Mc- grating clinical and laboratory data in genetic studies of com- Mahom FJ, McInnis MG, DePaulo JR Jr: Familial aggregation of plex phenotypes: a network-based data management system. psychotic symptoms in a replication set of 69 bipolar disorder Am J Med Genet 1998; 81:248–256 pedigrees. Am J Med Genet B Neuropsychiatr Genet 2003; 116: 22. Gibbons RD, Hedeker D: Applications of mixed-effect models in 90–97 biostatistics. Sankhya Series B 2000; 62:70–103 33. McInnis MG, McMahon FJ, Chase GA, Simpson SG, Ross CA, De- 23. Hedeker D, Gibbons RD: MIXREG: a computer program for Paulo JR Jr: Anticipation in bipolar affective disorder. Am J Hum mixed-effects regression analysis with autocorrelated errors. Genet 1993; 53:385–390 Comput Methods Programs Biomed 1996; 49:229–252 34. Angst J, Gamma A, Sellaro R, Lavori PW, Zhang H: Recurrence of 24. McElroy SL, Altshuler LL, Suppes T, Keck PE Jr, Frye MA, Denicoff bipolar disorders and major depression: a life-long perspec- KD, Nolen WA, Kupka RW, Leverich GS, Rochussen JR, Rush AJ, tive. Eur Arch Psychiatry Clin Neurosci 2003; 253:236–240 Post RM: Axis I psychiatric comorbidity and its relationship to historical illness variables in 288 patients with bipolar disorder. 35. Judd LL, Akiskal HS, Schettler PJ, Coryell W, Maser J, Rice JA, So- Am J Psychiatry 2001; 158:420–426 lomon DA, Keller MB: The comparative clinical phenotype and 25. Feinman JA, Dunner DL: The effect of alcohol and substance long term longitudinal episode course of bipolar I and II: a clin- abuse on the course of bipolar affective disorder. J Affect Dis- ical spectrum or distinct disorders? J Affect Disord 2003; 73:19– ord 1996; 37:43–49 32

1272 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Comparison of Rapid-Cycling and Non-Rapid-Cycling Bipolar Disorder Based on Prospective Mood Ratings in 539 Outpatients

Ralph W. Kupka, M.D., Ph.D. Objective: To detect risk factors for rapid number of lifetime manic or depressive ep- cycling in bipolar disorder, the authors isodes, history of rapid cycling, and history David A. Luckenbaugh, M.A. compared characteristics of rapid-cycling of drug abuse. The prevalence of these and non-rapid-cycling patients both from a characteristics increased progressively with Robert M. Post, M.D. categorical and a dimensional perspective. episode frequency. The proportion of Trisha Suppes, M.D., Ph.D. Method: Outpatients with bipolar I disor- women was greater than the proportion of der (N=419), bipolar II disorder (N=104), men only among patients with eight or Lori L. Altshuler, M.D. and bipolar disorder not otherwise speci- more episodes per year. The average time spent manic/hypomanic increased as a Paul E. Keck, Jr., M.D. fied (N=16) were prospectively evaluated with daily mood ratings for 1 year. Subjects function of episode frequency, but the av- Mark A. Frye, M.D. were classified as having rapid cycling erage time spent depressed was compara- (defined by the DSM-IV criterion of four or ble in patients with one episode and in Kirk D. Denicoff, M.D. more manic or depressive episodes within those with more than one episode. Brief episodes were as frequent as full-duration Heinz Grunze, M.D. 1 year) or not having rapid cycling, and the two groups’ demographic and retrospec- DSM-IV-defined episodes. Gabriele S. Leverich, M.S.W. tive and prospective illness characteristics Conclusions: A number of heteroge- were compared. Associated factors were neous risk factors were progressively asso- Susan L. McElroy, M.D. also evaluated in relationship to episode ciated with increasing episode frequency. frequency. Jörg Walden, M.D., Ph.D. Depression predominated in all bipolar Results: Patients with rapid cycling (N= disorder patients, but patients with rapid Willem A. Nolen, M.D., Ph.D. 206; 38.2%) significantly differed from cycling were more likely to be character- those without rapid cycling (N=333) with ized by manic features. The findings over- respect to the following independent vari- all suggest that rapid cycling is a dimen- ables: history of childhood physical and/or sional course specifier arbitrarily defined sexual abuse, bipolar I disorder subtype, on a continuum of episode frequency.

(Am J Psychiatry 2005; 162:1273–1280)

In 1974, Dunner and Fieve (1) defined patients with by the DSM-IV criteria. Subsequent studies gave further rapid cycling as having four or more manic or depressive support for the inclusion of brief episodes in the criteria episodes of at least 2 weeks’ duration in the year before the but provided little evidence in support of the boundary of study. Although subsequent studies addressed the issue of four episodes per year (3–6). rapid cycling from the perspectives of clinical and demo- In a meta-analysis of 20 clinical studies comparing graphic features, neurobiological dysfunction, longitudi- rapid-cycling and non-rapid-cycling bipolar disorder (6), nal course, and treatment response, it remains unclear to the overall prevalence of rapid cycling in unselected re- what extent rapid cycling delineates a distinct subtype or search samples was 16.3%, and rapid cycling was signifi- is merely an arbitrary point on a continuum of episode cantly associated with female gender and the bipolar II frequencies. disorder subtype. That analysis confirmed the need for The validity of rapid cycling as a course specifier for prospective evaluation in a large sample, particularly to DSM-IV bipolar I disorder and bipolar II disorder was re- address interrelationships among risk factors (3, 7). viewed by Bauer and Whybrow (2). They supported the To address these issues, we assessed episode frequency inclusion of rapid cycling as a course specifier despite from prospective daily mood ratings and compared retro- important unresolved issues, such as the paucity of char- spective and prospective illness variables in patients with acteristics that distinguish rapid-cycling patients (only fe- a rapid-cycling and a non-rapid-cycling course. In addi- male sex emerged from most studies), the arbitrariness of tion to classifying rapid cycling by using the traditional the criterion of four or more mood episodes per year, and definition of four or more mood episodes per year, we ex- the occurrence of brief but severe episodes that appeared amined these clinical characteristics in relationship to to be abundant in rapid cycling but were formally excluded the dimension of episode frequency in an attempt to dis-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1273 RAPID-CYCLING BIPOLAR DISORDER cern an empirically based boundary for rapid cycling, if ity of the mood episodes. Computer-calculated episode counts present. based on the DSM-IV criteria were compared to visually counted episodes from printed life charts in a subsample of 63 patients. The correlation between the two measures was significant for de- Method pressive episodes (r=0.77, p<0.0001) and for hypomanic, manic, and mixed episodes (r=0.99, p<0.0001). Subjects Data for all patients with at least 1 year of uninterrupted pro- Subjects were adult outpatients with DSM-IV bipolar I disorder spective daily Life-Chart Method ratings were included in the (N=419, 77.7%), bipolar II disorder (N=104, 19.3%), and bipolar dis- analysis. We excluded patients who left the Network before 1 year order not otherwise specified (N=16, 3.0%) diagnosed with the and patients with missing daily ratings at the time of the analysis. Structured Clinical Interview for DSM-IV (SCID) (8). All patients Patients with rapid cycling were defined as those with four or participated in the former Stanley Foundation Bipolar Network more depressive, hypomanic, manic, or mixed episodes (accord- with four sites in the United States (N=88, N=104, N=86, N=73), one ing to the DSM-IV duration criteria) in the first year of prospective in the Netherlands (N=143), and two in Germany (N=28, N=17), follow-up. and were enrolled from 1995 through 2000 (9). Outpatients were recruited from the participating clinics and in many cases were re- Data Analysis ferred by local physicians or were self-referred from local advocacy We compared demographic and retrospective and prospective groups; the only exclusion criterion was a current comorbid sub- illness characteristics of patients with DSM-IV-defined rapid stance use disorder that was severe enough to require treatment in cycling and patients without rapid cycling. We used chi-square a specialized setting (10). Interrater reliability for the diagnosis of analyses with Yates’s corrections for dichotomous variables and bipolar disorder was excellent (overall kappa of 0.92) (11). independent-sample t tests for continuous variables and applied Patients received treatment according to current standards in a Hochberg’s adjusted Bonferroni procedure for multiple tests of naturalistic fashion or in various more formal pharmacological significance (17). For the multivariate analysis, logistic regression treatment protocols also designed to match usual treatment. Pa- was used to examine significant independent contributions of the tients in this study group may have been included in studies or risk factors to rapid cycling. Variables that were significant in the clinical trials described in other published reports from the Stan- univariate analyses were included in the logistic regression analy- ley Foundation Bipolar Network. The study was approved by the sis. For risk factors that were highly intercorrelated, only one of institutional review boards of all participating sites, and all pa- the two factors was included in the model. Other risk factors were tients provided written informed consent. removed if multicollinearity was an issue in any given model. Variables that distinguished between patients with and with- Demographic and Clinical Characteristics out rapid cycling were also evaluated for their relationship with Patient- and clinician-rated questionnaires were used to gather episode frequency in the first year of prospective follow-up to re- data on demographic characteristics, functional status, prior veal discontinuities indicating a potential optimal cutoff point for illness characteristics, and psychiatric history of the parents as rapid cycling. Pearson’s product-moment correlations were used discussed elsewhere (9–12). The SCID was used to determine co- to supplement visual inspection of the curves for nonlinear rela- morbid axis I diagnoses. In addition, measures of thyroid param- tionships. All statistical analyses were performed with SPSS Ver- eters were available for a subgroup of 199 patients who simulta- sion 9.0 (SPSS, Inc., Chicago). neously participated in another study (13). Prospective Follow-Up and Episode Frequency Results The prospective course of illness was followed with the clini- Comparison of Patients cian-rated NIMH Life-Chart Method (14), which is used to record daily ratings of severity of manic and depressive symptoms on a With and Without Rapid Cycling 9-point graphic scale. Severity ratings were based on symptom- The 539 patients with 1 year of prospective ratings in- driven degree of functional impairment. At the baseline level (eu- cluded 237 men (44.0%) and 302 women (56.0%). Their thymia, score 0), there were neither significant mood symptoms mean age was 42.1 years (SD=11.5). A total of 206 (38.2%) nor functional impairment. Ratings for mania were as follows: 2.5=mild, 5=low moderate, 7.5=high moderate, and 10=severe. had rapid cycling, and 333 (61.8%) did not have rapid cy- Depression ratings were as follows: –2.5=mild, –5=low moderate, cling. A lifetime history of rapid cycling was reported at –7.5=high moderate, and –10=severe. A patient-rated Life-Chart study entry by 264 (50.7%) of 521 patients. The demo- Method was evaluated weekly to monthly by a research clinician, graphic and clinical characteristics of the current study who then completed the clinician-rated Life-Chart Method. The group did not differ from those of a larger group of 631 pa- validity and reliability of this procedure have been described elsewhere (15, 16). All clinicians were repeatedly trained and showed tients that included patients with a shorter follow-up pe- good interrater reliability (kappa=0.82) (15). riod or incomplete ratings (12). The number of episodes in 1 year was calculated by a computer The patients’ clinical characteristics at study entry are program by applying two sets of criteria to the data from the clini- summarized in Table 1. There was a nonsignificant over- cian-rated Life-Chart Method. First, in accordance with DSM-IV, representation of women in the rapid-cycling group. Edu- we identified episodes by using the following minimum criteria for severity and duration: 4 days of mild ratings for hypomania, 1 cational level, marital status, and work status did not differ week of moderate ratings or any hospitalization for mania, and 2 significantly between the groups (data not shown). Rapid weeks of moderate ratings for depression. An episode terminated cycling occurred in 41.3% of patients with bipolar I disor- with any switch to the opposite polarity or after 2 months of eu- der and 27.9% of patients with bipolar II disorder. Patients thymia. By using a second set of criteria, we also identified iso- with rapid cycling had an earlier age at onset, a longer du- lated severe episodes that lasted 1 or more days. The computer program calculated the number of days the patient experienced ration of illness, and a longer time from first symptoms to the various mood states and determined the mean level of sever- first medication treatment, compared to the patients with-

1274 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KUPKA, LUCKENBAUGH, POST, ET AL.

TABLE 1. Characteristics of Bipolar Disorder Patients With and Without Rapid Cycling in a 1-Year Prospective Study of Illness Coursea Patients With Patients Without Rapid Cycling Rapid Cycling Characteristic (N=206)b (N=333) Analysisc Mean SD Mean SD t df p Age (years) At study entry 42.4 11.1 42.5 11.9 0.11 537 0.91 At onset of first mood symptoms that affected functioning 17.6 9.1 23.1 10.0 5.98 462 <0.0001d At first depressive symptoms 18.9 9.9 23.9 10.5 5.17 462 <0.0001d At first manic/hypomanic symptoms 21.7 10.1 27.2 11.3 5.36 462 <0.0001d Time from first symptoms to first medication treatment (years) 11.2 10.3 6.8 8.9 4.73 450 <0.0001d Duration of illness from first mood symptoms (years) 24.3 11.8 19.4 11.6 4.39 462 <0.0001d Global Assessment of Functioning Scale score Past week before study entry 61.4 13.9 67.5 15.2 4.41 495 <0.0001d Best week in year before study entry 74.4 12.1 77.8 11.9 2.92 459 0.004 Worst week in year before study entry 43.5 15.1 48.7 17.2 3.24 457 0.001d

N % N % χ2 df p

Female gender 129 62.6 173 52.0 5.46 1 0.02 Bipolar disorder subtype 7.56 2 <0.03 Bipolar I disorder 173 84.0 246 73.9 Bipolar II disorder 29 14.1 75 22.5 Bipolar disorder not otherwise specified 4 1.9 12 3.6 Lifetime number of mood episodes at study entry 66.12 3 <0.0001d 1–4 5 2.6 38 12.8 5–10 12 6.2 75 25.3 11–20 2513.05618.9 >20 151 78.2 127 45.7 Rapid-cycling status Lifetime history 147 74.6 117 36.1 71.15 1 <0.0001d In last year before study entry 99 58.2 63 27.5 36.93 1 <0.0001d Psychotic mood episodes: lifetime history 114 61.3 173 55.6 1.31 1 0.25 Dysphoric mania/hypomania: lifetime history 137 71.7 153 48.1 26.19 1 <0.0001d One or more serious suicide attempts: lifetime history 70 35.7 82 26.8 4.09 1 <0.05 Exposure to antidepressants: lifetime history 187 93.5 268 83.8 9.82 1 0.002d Antidepressant-induced mania/hypomania: lifetime historye 107 61.1 118 49.0 5.58 1 0.018 Alcohol- or drug-induced mood episode: lifetime history 69 35.8 63 21.0 12.29 1 <0.0001d Thyroid status Lifetime history of hypothyroidismf 48 25.5 55 18.3 3.18 1 0.08 Positive thyroid autoantibody testg 30 33.3 27 24.8 1.37 1 0.24 a Number of subjects may differ slightly among variables because of missing data. b Rapid cycling was defined as four or more DSM-IV mood episodes recorded with the NIMH Life-Chart Method during the 1-year prospective follow-up. c Yates’s correction was used for chi-square tests for all two-by-two comparisons. d Statistically significant after Hochberg’s adjusted Bonferroni procedure for multiple tests. e In subjects with a lifetime exposure to antidepressants. f Most cases of clinical and subclinical hypothyroidism were currently or previously treated with thyroid hormone. g Data from 199 subjects also included in a previous study (13). out rapid cycling. There was a strong relationship between 6.49), bipolar I disorder subtype (Wald’s χ2=8.22, df=1, p= the occurrence of rapid cycling in any prior year, in the year 0.004, odds ratio=2.76), lifetime history of drug abuse before study entry, and during prospective follow-up. (Wald’s χ2=4.57, df=1, p=0.03, odds ratio=1.99), and history Rapid cycling was associated with a prior history of dys- of childhood physical and/or sexual abuse (Wald’s χ2= phoric mania/hypomania, lifetime treatment with antide- 4.67, df=1, p=0.03, odds ratio=1.86). pressants, and a history of substance-induced episodes. Rapid cycling was associated with lifetime DSM-IV anx- Prospective Course of Illness iety disorder, childhood physical or sexual abuse, and pa- As summarized in Table 3, patients with rapid cycling rental history of drug abuse, as detailed in Table 2. had a sevenfold greater mean number of full-duration Multivariate logistic regression (N=397) identified five manic and hypomanic episodes and a twofold greater independent variables that were significantly associated mean number of depressive episodes, compared to pa- with rapid cycling: a lifetime history of rapid cycling tients without rapid cycling. Over the course of the year, (Wald’s χ2=17.64, df=1, p<0.0001, odds ratio=3.39), higher patients with rapid cycling spent many more days in a number of lifetime mood episodes at study entry (11–20 manic or hypomanic state than did patients without rapid episodes: Wald’s χ2=5.24, df=1, p=0.02, odds ratio=4.66; cycling (73 days versus 25 days). Patients with rapid cy- >20 episodes: Wald’s χ2=8.51, df=1, p=0.004, odds ratio= cling also spent more days depressed, although the differ-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1275 RAPID-CYCLING BIPOLAR DISORDER

TABLE 2. Psychiatric Comorbidity, History of Early Abuse, and Parental Psychiatric History in Bipolar Disorder Patients With and Without Rapid Cycling in a 1-Year Prospective Study of Illness Coursea Patients With Rapid Cycling Patients Without Rapid Cycling (N=206)b (N=333) Analysisc Variable N % N % χ2 (df=1) p Lifetime diagnosis Substance abuse or dependence 93 45.4 121 36.4 3.84 0.05 Alcohol abuse or dependence 76 37.1 102 30.8 1.96 0.16 Drug abuse or dependence 56 27.3 57 17.2 7.26 0.007 Any anxiety disorder 102 50.2 101 30.7 19.51 <0.0001d Panic disorder 39 21.4 36 13.8 3.85 0.05 Posttraumatic stress disorder 12 8.5 7 3.6 2.87 0.09 History of physical and/or sexual abuse As a child 79 40.1 73 24.1 13.71 <0.0001d As an adolescent only 21 10.8 19 6.3 2.67 0.08 Parental history Major mood disorder 111 62.7 145 51.1 5.54 <0.02 Bipolar disorder 59 34.7 71 26.8 2.73 0.10 Depression 68 41.2 91 33.3 2.43 0.12 Suicide attempts 22 13.3 33 12.1 2.32 0.31 Substance abuse 67 35.6 67 22.3 9.61 0.002d Alcohol use disorder 57 30.3 64 21.4 4.45 <0.04 Drug abuse 24 15.3 10 3.5 19.09 <0.0001d a Number of subjects may differ slightly among variables because of missing data. b Rapid cycling was defined as four or more DSM-IV mood episodes recorded with the NIMH Life-Chart Method during the 1-year prospective follow-up. c Yates’s correction was used for all two-by-two comparisons. d Statistically significant after Hochberg’s adjusted Bonferroni procedure for multiple tests. ence between groups was smaller (145 days versus 121 one or more episodes per year. The number of additional days). During follow-up, patients with rapid cycling were brief episodes progressively increased along with the fre- twice as likely to experience dysphoric mania/hypomania quency of full-duration DSM-IV-defined episodes. as were patients without rapid cycling (82.5% versus 38.1%) Figure 2 shows the relationship between the overall fre- (χ2=99.56, df=1, p<0.0001). quency of full-duration DSM-IV-defined episodes and the Lithium, valproate, carbamazepine, or lamotrigine was prevalence of the variables that distinguished patients with used alone or in combination by 193 (93.7%) of the patients and without rapid cycling. Except for bipolar I disorder sub- with rapid cycling and 305 (91.6%) of the patients without type and female gender, these variables tended to increase rapid cycling. Patients with rapid cycling were more likely in prevalence as a function of episode number (r=0.15–0.37, than those without rapid cycling to be given antidepres- all p<0.0001). The prevalence of bipolar I disorder subtype sants (67% versus 57%) (χ2=4.60, df=1, p=0.03), antipsy- had a somewhat uneven pattern. An increase in the per- chotics (52% versus 32%) (χ2=19.08, df=1, p<0.00001), and centage of female patients at eight episodes could indicate thyroid hormone (32% versus 22%) (χ2=6.71, df=1, p=0.01). nonlinearity; 53% of patients with zero to seven episodes were women, compared to 73% of patients with eight or Variables Related to Episode Frequency more episodes (χ2=8.65, df=1, p=0.003), although the per- The subjects were grouped by the number of full-dura- centage of women among patients with nine and 10 or tion DSM-IV episodes in the first year of prospective fol- more episodes was lower than the percentage of women low-up, as follows: no episodes (N=74) and one (N=110), among patients with eight episodes. None of the other two (N=78), three (N=71), four (N=46), five (N=39), six (N= curves showed evidence of nonlinearity suggestive of this or 24), seven (N=24), eight (N=22), nine (N=13), and 10 or any other cutoff point for separating patients with rapid cy- more episodes (N=38). cling from those without rapid cycling. We reanalyzed the Figure 1 illustrates prospective illness characteristics in comparisons reported in Table 1 and Table 2 with a defini- relation to overall episode frequency. Increasing episode tion of rapid cycling as eight or more episodes; 73 subjects frequency was largely attributable to manic and hypomanic had rapid cycling according to this definition. Female gen- episodes. The number of DSM-IV depressive episodes was der was the only variable with a greater significance in the relatively constant in the groups of patients with five or reanalysis than in the original analysis (data not shown). more full-duration DSM-IV-defined mood episodes per year. Similarly, the average number of days per year during Discussion which patients experienced hypomanic, manic, or ultradian cycling gradually increased with the number of full- To our knowledge, this study is the first comparative in- duration DSM-IV-defined episodes per year, and the aver- vestigation of rapid-cycling and non-rapid-cycling bipolar age number of days during which patients were depressed disorder that systematically used both a categorical and a was relatively consistent among the groups of patients with dimensional analysis and is one of the few studies (18, 19)

1276 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KUPKA, LUCKENBAUGH, POST, ET AL.

TABLE 3. One-Year Prospective Illness Course in Bipolar Disorder Patients With and Without Rapid Cycling Patients With Rapid Cycling Patients Without Rapid Cycling (N=206)a (N=333)b Analysis Variable Mean SD Mean SD t df p Number of DSM-IV mood episodes All mood episodes 7.1 3.4 1.4 1.1 28.36 537 <0.0001c Manic episodesd 1.41.50.20.513.00537<0.0001c Hypomanic episodes 4.4 3.1 0.6 0.8 21.24 537 <0.0001c Depressive episodes 1.3 1.3 0.6 0.7 8.30 537 <0.0001c Number of additional brief mood episodese All brief mood episodes 8.9 9.6 2.3 3.4 11.45 537 <0.0001c Brief manic/hypomanic episodesd 4.55.31.12.110.60537<0.0001c Brief depressive episodes 4.4 5.7 1.3 1.9 10.26 537 <0.0001c Number of days in mood state Euthymia 121 78 215 110 10.38 504 <0.0001c Mania/hypomania 73 48 25 43 11.65 504 <0.0001c Hypomania 53 38 17 32 11.18 504 <0.0001c Moderate to severe mania 21 25 8 21 6.35 504 <0.0001c Ultradian cyclingf 26 47 3 10 8.27 504 <0.0001c Depression 145831211092.56504<0.02c Mild depression 65 43 64 62 0.26 504 0.79 Moderate to severe depression 79 69 57 81 3.11 504 0.002c Severity of mood episodesg Manic/hypomanic episodes 3.3 0.9 2.3 1.7 7.56 504 <0.0001c Depressive episodes –4.2 1.2 –3.4 1.5 6.48 504 <0.0001c a Rapid cycling was defined as four or more DSM-IV mood episodes recorded with the NIMH Life-Chart Method during the 1-year prospective follow-up. b Patients without rapid cycling included 31 euthymic subjects without any brief or full-duration mood episodes during the year. c Statistically significant after Hochberg’s adjusted Bonferroni procedure for multiple tests. d Manic episodes include episodes of euphoric and dysphoric (mixed) mania. e Brief mood episodes did not meet the DSM-IV duration criteria for a full-duration mood episode. f One or more distinct mood switches from depression to mania/hypomania or vice versa on the same day. g Severity of both full-duration and brief episodes was rated on a scale from 0 (euthymia) to 10 (severe) for mania and on a scale from 0 (euthymia) to –10 (severe) for depression. in which the diagnosis of rapid cycling was based on pro- and patients without rapid cycling were depressed 33.2% spectively observed episode frequency. of the time. It is interesting to note that in all patients (both In this group of 539 outpatients with bipolar disorder, with and without rapid cycling) who had one or more full- prospective evaluation revealed a 38% prevalence of rapid duration episodes per year, the average amount of time in cycling as defined by DSM-IV criteria. Notable findings in a depressed state was fairly stable, regardless of the total patients with rapid cycling were considerably greater se- number of episodes (Figure 1). In contrast, the proportion verity and frequency of manic episodes and a greater se- of time in a hypomanic, manic, or ultradian cycling state verity of depressive episodes, compared to patients with- was significantly higher in patients with rapid cycling out rapid cycling. Patients with rapid cycling had as many (27.1% of the time versus 7.7% of the time in patients with- brief mood episodes as full-duration DSM-IV episodes. out rapid cycling) and increased progressively as a func- However, there was a lack of clear boundaries between the tion of episode frequency. patients with and without rapid cycling on any of the pro- When we compared the prior illness histories of the pa- spective and retrospective variables examined as a tients with and without rapid cycling, we found that five function of episode number. independent variables were associated with rapid cycling The prospective course of illness indicates that the in the logistic regression analysis: previous rapid cycling, a study subjects were considerably ill despite extensive greater number of previous mood episodes, bipolar I dis- treatment. The prospective illness characteristics of the order subtype, history of childhood physical and/or sexual 419 bipolar I disorder patients were similar to findings in a abuse, and lifetime drug abuse. group of 146 patients with bipolar I disorder who were followed for 2–20 years in the Collaborative Depression Study A previous history of rapid cycling and a history of more (20). The patients in the Collaborative Depression Study than 10 mood episodes before study entry were the stron- were depressed an average of 31.9% of the time, manic for gest predictors of rapid cycling during the first year of pro- 2.3%, hypomanic for 7.0%, and cycling/mixed for 5.9%. spective follow-up. This finding suggests that some pa- The patients in our study experienced those mood states tients have a propensity toward rapid cycling that may be 35.6%, 3.9%, 8.7%, and 3.3% of the time, respectively. expressed periodically or more continuously. These findings confirm that the main burden for a ma- Our finding that rapid cycling was more prevalent among jority of patients with bipolar illness is depression (21); pa- patients with bipolar I disorder contrasts with various re- tients with rapid cycling were depressed 39.5% of the time, ports of overrepresentation of patients with rapid cycling

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1277 RAPID-CYCLING BIPOLAR DISORDER

FIGURE 1. Gender, Mood Episode Type, and Number of Days in Mood States in 539 Outpatients With Bipolar Disorder, by Frequency of Full-Duration DSM-IV Mood Episodes Prospectively Assessed in a 1-Year Study of Illness Coursea

Brief hypomania or mania Mania Euthymia Men Brief depression DSM-IV hypomania Hypomania Mild depression Women DSM-IV mania Ultradian Full depression DSM-IV depression cycling 120 28 360

100 24 300

20 80 240 16 60 180 12 40 120 8 Number of Subjects

20 4 60 Mean Number of Days per Year

0 Mean Number of Episodes per Year 0 0 0123456789≥10 0123456789≥10 0123456789≥10

Number of Prospectively Assessed DSM-IV Mood Episodes per Year a The number of subjects in each DSM-IV mood episode frequency group was as follows: no episodes (N=74) and one (N=110), two (N=78), three (N=71), four (N=46), five (N=39), six (N=24), seven (N=24), eight (N=22), nine (N=13), and 10 or more episodes (N=38).

FIGURE 2. Prevalence of Risk Factors for Rapid Cycling in 539 Patients With Bipolar Disorder, by Frequency of Full-Duration DSM-IV Mood Episodes Prospectively Assessed in a 1-Year Study of Illness Coursea

Lifetime history of Lifetime history of Lifetime anxiety disorder Female rapid cycling exposure to antidepressants Lifetime history of more than 20 mood Lifetime history of Physical and/or sexual Bipolar I disorder episodes at study entry antidepressant-induced abuse as a child subtype mania/hypomania Lifetime history of dysphoric mania/ Lifetime history of Lifetime drug abuse or hypomania alcohol- or drug-induced dependence mood episode 100

40 Prevalence (%) Prevalence

0 02468≥10 0 2 4 6 8 ≥1002468≥1002468≥10 Number of Prospectely Assessed DSM-IV Mood Episodes per Year a The number of subjects in each DSM-IV mood episode frequency group was as follows: no episodes (N=74) and one (N=110), two (N=78), three (N=71), four (N=46), five (N=39), six (N=24), seven (N=24), eight (N=22), nine (N=13), and 10 or more episodes (N=38). among bipolar II disorder patients (6). However, two large drug abuse and anxiety disorders (12). These findings sug- studies also failed to report a preponderance of rapid- gest that early traumatic experiences may contribute to a cycling patients among patients with bipolar II disorder later adverse illness course. (22, 23). Substance abuse is highly comorbid with bipolar disor- Childhood physical or sexual abuse was associated with der (11), and patients with rapid cycling may be particu- higher episode frequency. We previously reported that larly sensitive to the destabilizing properties of alcohol early abuse was associated with an earlier age at onset of and drugs, as they frequently reported prior induction of bipolar disorder, serious suicide attempts, and comorbid depressive and manic/hypomanic episodes by these agents.

1278 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KUPKA, LUCKENBAUGH, POST, ET AL.

Alternatively, patients with faster cycling frequencies may racies. An inevitable limitation is the fact that we studied use these substances at a higher rate. the naturalistic course of treated bipolar disorder. Several factors were associated with rapid cycling in the Despite these limitations, our findings can be used in univariate but not in the multivariate analysis, indicating the further conceptualization of rapid cycling in bipolar that these factors were intercorrelated with one or more of disorder, defined as the occurrence of four or more mood the factors mentioned earlier. As in other studies (6), we episodes per year (1). Kendell (29) proposed that bound- found a modest overrepresentation of women among the aries between two clinical syndromes may be indicated by patients with rapid cycling, especially among those with a nonlinear relationship between symptom measures and higher episode frequencies. Our finding of a younger age an independent variable on which the syndromes differ at onset of bipolar disorder in patients with rapid cycling markedly. Like Bauer et al. (3), we found no bimodal distri- was also found in previous studies (23–25), as was our bution of the number of episodes per year. They reported finding of a longer duration of illness in patients with that the proportion of women increased markedly among rapid cycling (4, 5), although most studies reviewed found patients with four to eight episodes per year and even no differences (6). Our finding of a longer time between reached 100% among patients with nine or more episodes first symptoms and first medication treatment in patients per year. We found a larger proportion of women among with rapid cycling raises the question of whether earlier patients with eight or more episodes per year. However, recognition and treatment might prevent or attenuate this this potential nonlinearity was not confirmed in analyses more problematic course. of any of the other risk factors we examined (Figure 2). Prior histories of both dysphoric mania/hypomania and Likewise, changing the boundary for rapid cycling from ultradian cycling were more prevalent in patients with four to eight episodes per year did not essentially change rapid cycling than in patients without rapid cycling. These the results of our dichotomous comparisons. associations were confirmed during prospective follow- In conclusion, our overall findings suggest that rapid up, which suggests that depressive features pervade both cycling is a dimensional course specifier with an arbi- manic and depressive phases of the illness in patients with trarily defined cutoff at four episodes per year on a contin- faster cycle frequencies. Our study was not designed to uum of episode frequency seen in the naturalistic course evaluate the contribution of antidepressants to the devel- of treated outpatients with bipolar disorder. opment of rapid cycling. Still, 54% of patients in the study who had been treated with antidepressants reported hav- Presented at the Fifth International Conference on Bipolar Disor- ing experienced an antidepressant-induced switch to ma- der, Pittsburgh, June 12–14, 2003. Received Feb. 19, 2004; revisions received May 14 and June 29, 2004; accepted Aug. 2, 2004. From Al- nia/hypomania in the past, a phenomenon that has been trecht Institute for Mental Health Care and University Medical Center associated with a subsequent rapid-cycling course (26). Utrecht, Utrecht, Netherlands; the Mood and Anxiety Disorders Pro- The most rigorous studies of family history of mood dis- gram and the Biological Psychiatry Branch, NIMH, Bethesda, Md.; the University of Texas Southwestern Medical Center, Dallas; the UCLA orders reported no significant differences in family history Mood Disorders Research Program, Los Angeles; the Psychopharma- between patients with and without rapid cycling (4–6, 19, cology Research Program, Department of Psychiatry, University of 27, 28). We found that a parental history of major mood Cincinnati College of Medicine, Cincinnati; the Mental Health Care Line and General Clinical Research Center, Cincinnati Veterans Affairs disorder (bipolar disorder and/or depression) was more Medical Center, Cincinnati; Psychiatrische Klinik der Ludwig-Maximil- likely in patients with rapid cycling than in those without ian Universität, Munich; Universitätsklinik für Psychiatrie der Univer- rapid cycling. More prominent was our finding of a paren- sität Freiburg, Freiburg, Germany; and the Department of Psychiatry, University of Groningen, Groningen, Netherlands. Address correspon- tal history of drug abuse in patients with rapid cycling. dence and reprint requests to Dr. Kupka, Altrecht Institute for Mental However, rapid cycling was also associated with the occur- Health Care, Tolsteegsingel 2A, 3582 AC Utrecht, Netherlands; rence of childhood adversity, which suggests that parental [email protected] (e-mail). Supported by the Stanley Medical Research Institute. substance abuse and the associated environmental insta- The authors thank Adriaan Honig, M.D., Ph.D., Baer Arts, M.D., Titus bility may have indirectly contributed to the risk of rapid van Os, M.D., Ph.D., Pieternel Kölling, M.D., Herro Kraan, M.D., Ph.D., cycling. Max Sonnen, M.D., Rocco Hoekstra, M.D., Onno Habekotté, M.D., Harm-Jan Pot, M.D., Albert Blom, M.D., and Rein Holleboom, M.D., The results of our study must be considered in the con- for patient recruitment and data collection in the Netherlands. text of several limitations. Although our study population is in many respects comparable to other groups of outpatients (10), 80% of the subjects in our study had bipolar I References disorder, almost 60% reported a lifetime history of psy- 1. Dunner DL, Fieve RR: Clinical factors in lithium carbonate pro- chotic symptoms, and 38% had a rapid-cycling course. phylaxis failure. Arch Gen Psychiatry 1974; 30:229–233 This greater overall severity of illness may restrict the gen- 2. Bauer MS, Whybrow P: Validity of rapid cycling as a modifier eralizability of our findings. The exclusion of patients with for bipolar disorder in DSM-IV, in DSM-IV Source Book, vol 2. less than 1 year of follow-up may have affected the repre- Edited by Widiger TA, Frances AJ, Pincus HA, Ross R, First MB, Davis WW. Washington, DC, American Psychiatric Association, sentativeness of the study subjects. Moreover, the retro- 1996, pp 299–314 spective data on previous illness characteristics covered a 3. Bauer MS, Calabrese J, Dunner DL, Post R, Whybrow PC, Gyulai period of many years and thus were susceptible to inaccu- L, Tay LK, Younkin SR, Bynum D, Lavori P, Price AR: Multisite

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1279 RAPID-CYCLING BIPOLAR DISORDER

data reanalysis of the validity of rapid cycling as a course mod- Method (NIMH-LCM-p) for longitudinal assessment of bipolar ifier for bipolar disorder in DSM-IV. Am J Psychiatry 1994; 151: illness. Psychol Med 2000; 30:1391–1397 506–515 16. Meaden PM, Daniels RE, Zajecka J: Construct validity of life 4. Maj M, Magliano L, Pirozzi R, Marasco C, Guarneri M: Validity of chart functioning scales for use in naturalistic studies of bipo- rapid cycling as a course specifier for bipolar disorder. Am J lar disorder. J Psychiatr Res 2000; 34:187–192 Psychiatry 1994; 151:1015–1019 17. Hochberg Y: A sharper Bonferroni procedure for multiple tests 5. Maj M, Pirozzi R, Formicola A, Tortorella A: Reliability and va- of significance. Biometrika 1988; 75:800–802 lidity of four alternative definitions of rapid-cycling bipolar dis- 18. Cowdry RW, Wehr TA, Zis A, Goodwin FK: Thyroid abnormalities order. Am J Psychiatry 1999; 156:1421–1424 associated with rapid cycling bipolar illness. Arch Gen Psychia- 6. Kupka RW, Luckenbaugh DA, Post RM, Leverich GS, Nolen WA: try 1983; 40:414–420 Rapid and non-rapid cycling bipolar disorder: a meta-analysis 19. Coryell W, Endicott J, Keller M: Rapidly cycling affective disor- of clinical studies. J Clin Psychiatry 2003; 64:1483–1494 der: demographics, diagnosis, family history, and course. Arch 7. Goodwin FK, Jamison KR: Manic-Depressive Illness. New York, Gen Psychiatry 1992; 49:126–131 Oxford University Press, 1990 20. Judd LJ, Akiskal HS, Schettler PJ, Endicott J, Maser J, Solomon DA, Leon AC, Rice JA, Keller MB: The long-term natural history 8. First MB, Spitzer RL, Gibbon M, Williams JBW: Structured Clini- of the weekly symptomatic status of bipolar I disorder. Arch cal Interview for DSM-IV Axis I Disorders, Research Version, Pa- Gen Psychiatry 2002; 59:530–537 tient Edition, version 2.0 (SCID-I/P). New York, New York State 21. Calabrese JR, Shelton MD, Bowden CL, Rapport DJ, Suppes T, Psychiatric Institute, Biometrics Research, 1996 Shirley ER, Kimmel SE, Caban SJ: Bipolar rapid cycling: depres- 9. Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck PE, Denicoff sion as its hallmark. J Clin Psychiatry 2001; 62(suppl 14):34–41 KD, Suppes T, Altshuler LL, Kupka R, Kramlinger KG, Post RM: 22. Serretti A, Mandelli L, Lattuada E, Smeraldi E: Rapid cycling The Stanley Foundation Bipolar Treatment Outcome Network, mood disorder: clinical and demographic features. Compr Psy- I: longitudinal methodology. J Affect Disord 2001; 67:33–44 chiatry 2002; 43:336–343 10. Suppes T, Leverich GS, Keck PE, Nolen WA, Denicoff KD, Alt- 23. Schneck CD, Miklowitz DJ, Calabrese JR, Allen MH, Thomas MR, shuler LL, McElroy SL, Rush AJ, Kupka R, Frye MA, Bickel M, Post Wisniewski SR, Miyahara S, Shelton MD, Ketter TA, Goldberg JF, RM: The Stanley Foundation Bipolar Treatment Outcome Net- Bowden CL, Sachs GS: Phenomenology of rapid-cycling bipolar work, II: demographics and illness characteristics of the first disorder: data from the first 500 participants in the Systematic 261 patients. J Affect Disord 2001; 67:45–49 Treatment Enhancement Program. Am J Psychiatry 2004; 161: 11. McElroy SL, Altshuler LL, Suppes T, Keck PE Jr, Frye MA, Denicoff 1902–1908 KD, Nolen WA, Kupka RW, Leverich GS, Rochussen JR, Rush AJ, 24. Bowden CL, Calabrese JR, McElroy SL, Rhodes LJ, Keck PE Jr, Post RM: Axis I psychiatric comorbidity and its relationship to Cookson J, Anderson J, Bolden-Watson C, Asher J, Monaghan E, historical illness variables in 288 patients with bipolar disorder. Zhou J: The efficacy of lamotrigine in rapid cycling and non- Am J Psychiatry 2001; 158:420–426 rapid cycling patients with bipolar disorder. Biol Psychiatry 12. Leverich GS, McElroy SL, Suppes T, Keck PE Jr, Denicoff KD, 1999; 45:953–958 Nolen WA, Altshuler LL, Rush AJ, Kupka R, Frye MA, Autio KA, 25. Carter TD, Mundo E, Parikh SV, Kennedy JL: Early age at onset Post RM: Early physical and sexual abuse associated with an as a risk factor for poor outcome of bipolar disorder. J Psychiatr adverse course of bipolar illness. Biol Psychiatry 2002; 51:288– Res 2003; 37:297–303 297 26. Koukopoulos A, Caliari B, Tundo A, Minnai G, Floris G, Reginaldi 13. Kupka RW, Nolen WA, Post RM, McElroy SL, Altshuler LL, Deni- D, Tondo L: Rapid cyclers, temperament, and antidepressants. coff KD, Frye MA, Keck PE Jr, Leverich GS, Rush AJ, Suppes T, Pol- Compr Psychiatry 1983; 24:249–258 lio C, Drexhage HA: High rate of autoimmune thyroiditis in bi- 27. Nurnberger J Jr, Guroff JJ, Hamovit J, Berrettini W, Gershon E: A polar disorder: lack of association with lithium exposure. Biol family study of rapid-cycling bipolar illness. J Affect Disord Psychiatry 2002; 51:305–311 1988; 15:87–91 14. Leverich GS, Post RM: Life charting of affective disorders. CNS 28. Lish JD, Gyulai L, Resnick SM, Kirtland A, Amsterdam JD, Why- Spectr 1998; 3:21–37 brow PC, Price AR: A family history study of rapid-cycling bipo- 15. Denicoff KD, Leverich GS, Nolen WA, Rush AJ, McElroy SL, Keck lar disorder. Psychiatry Res 1993; 48:37–46 PE, Suppes T, Altshuler LL, Kupka R, Frye MA, Hatef J, Brotman 29. Kendell RE: The choice of diagnostic criteria for biological re- MA, Post RM: Validation of the prospective NIMH Life-Chart search. Arch Gen Psychiatry 1982; 39:1334–1339

1280 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Olanzapine Versus Lithium in the Maintenance Treatment of Bipolar Disorder: A 12-Month, Randomized, Double-Blind, Controlled Clinical Trial

Mauricio Tohen, M.D., Dr.P.H. Robert W. Baker, M.D. nia Rating Scale or Hamilton depression scale) occurred in 30.0% of olanzapine- Waldemar Greil, M.D. Heidi Crane, M.S. treated and 38.8% of lithium-treated patients. The noninferiority of olanzapine Joseph R. Calabrese, M.D. Martin R. Dossenbach, M.D. relative to lithium (primary objective) in preventing relapse/recurrence was met, Gary S. Sachs, M.D. Charles L. Bowden, M.D. since the lower limit of the 95% confi- Lakshmi N. Yatham, M.B., dence interval on the 8.8% risk difference (–0.1% to 17.8%) exceeded the predefined F.R.C.P.C. Objective: The authors compared the ef- noninferiority margin (–7.3%). Secondary ficacy of olanzapine and lithium in the results showed that compared with lith- Bruno Müller Oerlinghausen, prevention of mood episode relapse/ Dr.Med. ium, olanzapine had significantly lower recurrence. risks of manic episode and mixed episode Athanasios Koukopoulos, M.D. Method: Patients with a diagnosis of bi- relapse/recurrence. Depression relapse/ polar disorder (manic/mixed), a history of recurrence occurred in 15.7% of olanza- Giovanni B. Cassano, M.D. two or more manic or mixed episodes pine-treated and 10.7% of lithium-treated within 6 years, and a Young Mania Rating patients. Mean weight gain during open- ≥ Heinz Grunze, M.D. Scale total score 20 entered the study label cotreatment was 2.7 kg; during dou- and received open-label cotreatment ble-blind monotherapy, weight gain was Rasmus W. Licht, M.D., Ph.D. with olanzapine and lithium for 6–12 significantly greater with olanzapine (1.8 weeks. Those meeting symptomatic re- kg) than with lithium (–1.4 kg). Liliana Dell’Osso, M.D. mission criteria (Young Mania Rating Scale score ≤12; 21-item Hamilton de- Conclusions: These results suggest that Angela R. Evans, Ph.D. pression scale score ≤8) were randomly olanzapine was significantly more effec- assigned to 52 weeks of double-blind tive than lithium in preventing manic and Richard Risser, M.Sc. monotherapy with olanzapine, 5–20 mg/ mixed episode relapse/recurrence in pa- day (N=217), or lithium (target blood tients acutely stabilized with olanzapine level: 0.6–1.2 meq/liter) (N=214). and lithium cotreatment. Both agents Results: Symptomatic relapse/recur- were comparable in preventing depres- rence (score ≥15 on either the Young Ma- sion relapse/recurrence.

(Am J Psychiatry 2005; 162:1281–1290)

Despite ongoing maintenance therapy, patients with lamotrigine (7), and carbamazepine (8) also have been bipolar disorder will experience frequent fluctuations in used as maintenance therapies. symptom severity and multiple relapses. Prospective Olanzapine has shown superiority to placebo in treating naturalistic studies examining relapse have reported re- acute manic episodes in patients with bipolar I disorder lapse risks ranging from 44% in 1 year (1) to 73%–88.7% (9, 10). Furthermore, in a 47-week comparative trial of over 4–5 years (2, 3). To date, a limited number of thera- olanzapine versus valproate, rates of mood episode fol- peutic agents have been available for the long-term treat- lowing acute remission of mania were comparable be- ment of bipolar disorder. Lithium has been the mainstay tween the groups (11), suggesting that olanzapine may be of maintenance therapy for >30 years. It is the most exten- effective in the prevention of bipolar disorder relapse/re- sively studied mood stabilizer and has the best overall effi- currence. cacy for the prophylactic treatment of bipolar disorder. This trial compared the efficacy of olanzapine and lith- Recent meta-analyses show it to be superior to placebo in ium for the prevention of mood episode relapse/recur- the prevention of relapse (4) and to reduce the risk of re- rence. For simplicity, the term recurrence will be used lapse 3.6-fold (5). The anticonvulsant agents valproate (6), throughout the text.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1281 OLANZAPINE VERSUS LITHIUM IN BIPOLAR MAINTENANCE

TABLE 1. A Priori Categorical Definitions of Remission Dur- the double-blind taper period, patients remained on their current ing Olanzapine and Lithium Cotreatment and Recurrence dose of randomly assigned treatment, and the dose of the discon- Following Random Assignment to Double-Blind Olanzapine tinued drug was tapered in a blinded, a priori-determined, step- or Lithium Monotherapy in Patients With Bipolar Disorder wise manner over 4 weeks. Categorical Lithium levels were monitored every 2 weeks during the dou- Definition Remission Recurrence ble-blind taper period and monthly during double-blind mainte- Symptomatic Young Mania Rating Scale Score ≥15 on Young nance monotherapy. If the serum level of lithium deviated from total ≤12; and Hamilton Mania Rating the therapeutic range during these study periods, the investigator ≤ depression scale score 8 Scale and/or was to adjust the dose of lithium to reestablish blood levels within Hamilton the therapeutic range, with a goal of reaching this range within 30 depression scale days. Serum levels ranging from 0.6–1.2 meq/liter were consid- Syndromic Mania: All DSM-IV A and B Meeting DSM-IV criteria for current manic criteriaa for ered within normal limits. Maintenance of the blind associated episode no worse than mild current manic, with blood draws has been described (12). Briefly, all patients ran- (≤3 on a scale of 1–7) and no depressive, or domly assigned to olanzapine also had blood drawn. For every more than two B criteria mixed episode outlier report generated for a lithium patient, a sham lithium out- given mild rating (3 on a scale lier report was sent to an olanzapine patient. Thus, reports to in- of 1–7) vestigative sites indicating that the lithium dose should be ad- Depression: All DSM-IV A justed did not unmask the randomized assignment. criteria for current major depressive episode no worse Concomitant Medications than mild (≤3 on a scale of 1–7) and no more than three Patients who entered the study receiving psychotropic medi- A criteria given mild rating cations (including anticonvulsants, typical or atypical antipsy- (3 on a scale of 1–7) chotics [oral or intramuscular], or antidepressants) were gradu- a Not including duration criterion. ally discontinued from these medications at the discretion of the investigator during the first 3 weeks of the open-label cotreatment period. However, oral or intramuscular haloperidol and Method zuclopenthixol were permitted for extreme agitation during the open-label period. Benzodiazepines were allowed according to Patients the following guidelines. The maximum dose from the screening Patients enrolled in this study were ≥18 years of age and met period through the first 6 weeks of the open-label cotreatment DSM-IV criteria for bipolar disorder (current episode manic or period was 8 mg/day in lorazepam equivalents and 6 mg/day in mixed) as determined with the Structured Clinical Interview for lorazepam equivalents for the remainder of the open-label pe- DSM-IV, Patient Version. Patients were required to have a Young riod and during the first 2 weeks of the taper period. It was fur- Mania Rating Scale total score ≥20 at baseline and a history of at ther decreased to 4 mg/day for the remaining 2 weeks of the least two manic or mixed episodes in the preceding 6 years. Pa- taper period and then to 2 mg/day (for not more than 60 cumu- tients were excluded from the study if they had a serious, unstable lative days) for the double-blind monotherapy period. Patients medical illness; met DSM-IV substance dependence criteria (nico- were permitted concomitant medication for treatment-emer- tine or caffeine excepted) within the past 30 days; had been treated gent extrapyramidal symptoms (biperiden or benztropine mesy- ≤ ≤ with a depot neuroleptic within 6 weeks of random assignment; or late, 6 mg/day; trihexyphenidyl, 12 mg/day). However, pro- were considered a serious suicide risk. Patients were also excluded phylactic use of anticholinergics for extrapyramidal symptoms if they had a history of intolerance, or lack of response, to an ade- was not allowed. quate trial of lithium or olanzapine as determined by the investi- Assessments gator. After the study was completely described to the patients, written informed consent was obtained. The study was approved Recurrence and severity of illness were assessed with the Young by the appropriate ethics review boards. Mania Rating Scale and the 21-item Hamilton depression scale; raters were trained and certified to use these scales. A minimum Study Design reliability score (intraclass correlation [ICC]) of 0.75 was required Patients were recruited from 87 inpatient and outpatient set- for certification; raters who failed to achieve an ICC of ≥0.75 were tings across Western Europe, Canada, South Africa, Israel, Austra- retrained and tested again. Four hundred fifty-two raters were lia, and New Zealand between August 1999 and June 2002. This certified to use the Young Mania Rating Scale and Hamilton de- randomized, double-blind, controlled trial consisted of four pression scale for this study, with average ICCs of 0.88 (Hamilton study periods: 1) screening (two clinic visits over 2–7 days), 2) depression scale) and 0.90 (Young Mania Rating Scale). The vast open-label cotreatment (6–12 weeks; twice-weekly visits for the majority of raters achieved an ICC of ≥0.85 (78.8% for the Hamil- first 2 weeks, weekly thereafter), 3) double-blind taper (4 weeks; ton depression scale and 77.4% for the Young Mania Rating weekly visits), and 4) double-blind monotherapy (48 weeks; bi- Scale). Patient safety was evaluated through standard clinical ob- weekly during the first 4 weeks, monthly thereafter). Eligible pa- servations, and extrapyramidal symptoms were assessed with the tients began open-label cotreatment with olanzapine, 15 mg/ Simpson-Angus Rating Scale, the Barnes Rating Scale for Drug-In- day, and lithium, 600 mg/day. Allowed dosages of olanzapine duced Akathisia, and the Abnormal Involuntary Movement Scale. were 5–20 mg/day. Investigators were required to optimize lith- Criteria for treatment-emergent extrapyramidal symptoms have ium dose and reach a target blood level of 0.6–1.2 meq/liter by been described previously (12). week 4 during this period. Statistical Methods Patients who met symptomatic remission criteria (Table 1) during the open-label cotreatment period were randomly reassigned The primary objective of the study was the assessment of olan- in a 1:1 ratio by means of a unique drug kit number (via a call-in zapine’s noninferiority to lithium in the risk of symptomatic mood Interactive Voice Response System) to monotherapy with either episode recurrence. It was estimated that it would require that 200 olanzapine or lithium. All patients, study site personnel, and patients in symptomatic remission be randomly assigned to each sponsor investigators were blind to randomization codes. During therapy to provide 80% power to detect the protocol-defined mar-

1282 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 TOHEN, GREIL, CALABRESE, ET AL.

TABLE 2. Treatment Discontinuation Rates Among Bipolar Disorder Patients During Open-Label Acute Cotreatment With Olanzapine and Lithium and Double-Blind Olanzapine or Lithium Monotherapy

Open-Label Acute Double-Blind Maintenance Monotherapy Cotreatment With (N=431) Olanzapine and Lithium Olanzapine Lithium b b (N=543) (N=217) (N=214) Fisher’s exact Reason for Discontinuationa N% N%N% p Adverse events 34 6.3 41 18.9 55 25.7 0.11 Lack of efficacy 16 2.9 31 14.3 34 15.9 0.69 Patient decisionc 37 6.8 28 12.9 34 15.9 0.42 Criteria not met/noncompliance 8 1.5 12 5.5 11 5.1 1.00 Sponsor decision 1 0.2 0 0.0 1 0.5 0.50 Investigator decision 9 1.7 2 0.9 8d 3.7 0.07 Lost to follow-up 2 0.9 1 0.5 1.00 a Based on the investigator’s choice of a single primary reason for ending participation in the study. b Number included in analysis of primary outcome. c Includes patients who chose to discontinue due to perception of satisfactory response. d For four of these patients, the investigator listed the reason as “due to noncompliance.” gin of noninferiority for the reduction in absolute risk for olanza- trial (46.5% [N=101] versus 32.7% [N=70], respectively; p= pine relative to lithium: a 95% confidence interval lower limit at or 0.004, Fisher’s exact test). The most common reasons for below –0.073 (–7.3%). The calculation assumed an expected risk discontinuation during the double-blind maintenance pe- difference of 0.045 (0.319 for olanzapine, 0.364 for lithium). Differ- ences in recurrence risk were tested by using Fisher’s exact test, riod were adverse events, lack of efficacy, and patient deci- and the 95% confidence interval about the absolute risk reduction sion (Table 2). There were no significant between-group was computed using the normal approximation. Odds ratios with differences in reasons for premature discontinuation. The 95% confidence intervals are also presented. estimated median time to discontinuation was 303 and 207 Categorical baseline characteristics (as well as adverse event in- days for olanzapine- and lithium-treated patients, respec- cidence rates and other categorical outcomes) were compared be- tively, and the time to discontinuation for any reason was tween treatments with Fisher’s exact test. For continuous baseline χ2 data, Wilcoxon rank-sum tests or analyses of variance (ANOVAs) significantly earlier for patients receiving lithium ( =5.7, were used. ANOVA models (with terms for treatment, country, df=1, p<0.02, log-rank test). Because symptomatic recur- and the treatment-by-country interaction) were also used to ana- rence in a few cases did not necessarily result in immediate lyze the last observation carried forward baseline-to-endpoint discontinuation from the trial, a post hoc tabulation of the changes for the continuous safety measures (laboratory analytes, number of completers with sustained remission through- vital signs, and weight). Time-to-event data were analyzed by using Kaplan-Meier estimated survival curves with comparisons out the double-blind phase of the study also shows signifi- made by using log-rank tests. A logistic regression analysis was cantly higher completion with sustained remission for performed to examine the impact of quantitative length of re- olanzapine (43.3% [N=94 of 217]) than lithium (28.5% [N= mission before treatment assignment on the difference between 61 of 214]) (p=0.002, Fisher’s exact test). If those patients treatments in mood episode recurrence risk (test of interaction who had a recurrence were also counted as completers, between treatment and length of remission). Analyses were completed on an intent-to-treat basis; SAS software version 6 (Cary, then completion rates were 73.3% for olanzapine and N.C., SAS Institute) was used to perform all analyses; treatment 67.3% for lithium (p=0.206, Fisher’s exact test). effects were tested at the two-sided alpha level of 0.05 and inter- Patient characteristics are presented in Table 3. Approxi- actions at a level of 0.10 as the protocol specified. mately 93% of the patients had a manic index episode, and 26% were experiencing psychotic features. Among patients Results randomly assigned to a treatment condition, 72.2% were hospitalized for treatment of their index episode at the Patient Disposition and Characteristics time they entered into the open-label cotreatment phase A total of 543 patients were enrolled during the open-la- (lithium=74.8%; olanzapine=69.6%) (p=0.24, Fisher’s exact bel period and received lithium/olanzapine cotreatment. test). Overall, treatment groups were comparable with re- Ultimately, 431 (79.4%) achieved symptomatic (protocol- spect to demographic and clinical characteristics. defined) remission criteria during the open-label period The mean doses of olanzapine and lithium, respectively, and were randomly assigned to double-blind maintenance during the open-label period were 13.5 mg/day (SD=4.0) monotherapy with olanzapine (N=217) or lithium (N=214). and 1003.3 mg/day (SD=267.0) (mean serum level=0.697 The most common reasons for discontinuation during the meq/liter, SD=0.14). During the open-label phase, investi- open-label period were patient decision and adverse gators were to titrate the lithium dose to attain therapeutic events (Table 2). Of the patients achieving symptomatic re- levels by week 4. Considering the post-titration stabiliza- mission, 91.6% (N=395) achieved syndromic remission. tion period only (from week 4 to random assignment), the One hundred seventy-one patients completed the double- mean dose of lithium was 1097.0 mg (SD=277.0), and the blind maintenance period, with significantly more olanza- mean lithium serum level was 0.76 meq/liter (SD=0.14). pine-treated than lithium-treated patients completing the For the double-blind period, the mean dose was 11.9 mg

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1283 OLANZAPINE VERSUS LITHIUM IN BIPOLAR MAINTENANCE

TABLE 3. Demographic and Clinical Characteristics of Bipolar Disorder Patients Stabilized With Olanzapine and Lithium Cotreatment Then Randomly Assigned to Double-Blind Olanzapine or Lithium Monotherapy

Open-Label Acute Cotreatment Double-Blind Maintenance Monotherapy With Olanzapine and Lithium Olanzapine Lithium Characteristic (N=543) (N=217) (N=214) Analysis N % N % N % pa

Female 290 53.4 113 52.1 115 53.7 0.77 Caucasian 539 99.3 214 98.6 214 100 0.25 Manic index episode 503 92.6 202 93.1 202 94.4 0.69 Psychotic features present 149 27.4 59 27.2 53 24.8 0.58 History of rapid cycling course 22 4.1 6 2.8 7 3.3 0.96 Lifetime psychotropic medication use Lithium 401 73.8 161 74.2 160 74.8 0.91 Antipsychotic 509 93.7 202 93.1 204 95.3 0.41 Valproate 149 27.4 61 28.1 52 24.3 0.38

Interquartile Interquartile Median Range Median Range pb Number of lifetime mood episodes Mania 4 4 4 5 0.77 Depression 2 2 2 3 0.16

Mean SD Mean SD Mean SD F df pc

Age (years) 42.9 13.0 42.5 13.1 42.3 12.3 0.21 1, 387 0.65 Length of current episode (days) 37.5 37.9 37.7 39.2 37.0 33.3 0.60 1, 386 0.44 Time in remission before randomization (days) 19.7 19.0 20.6 19.3 0.39 1, 387 0.53 Young Mania Rating Scale total scored 25.8 7.2 3.6 3.5 3.9 3.8 1.8 1, 387 0.18 Hamilton depression scale total scored 5.7 4.7 1.6 2.0 1.6 1.9 1.0 1, 387 0.31 a Fisher’s exact test. b Wilcoxon rank-sum test. c ANOVA. d The mean Young Mania Rating Scale and Hamilton depression scale scores at screening were calculated from 533 patients; 10 patients who underwent initial screening did not have baseline Young Mania Rating Scale and Hamilton depression scale total scores.

TABLE 4. Mean Serum Lithium Levels During Double-Blind Incidence of and Time to Mood Episodes Maintenance Monotherapy With Olanzapine or Lithium in Bipolar Disorder Patients Following Stabilization With Symptomatic recurrence of any mood episode following Olanzapine and Lithium Cotreatment remission of mania or depression was observed in 38.8% Lithium Serum Level of lithium-treated and 30.0% of olanzapine-treated pa- (meq/liter) tients (Table 5). Statistical noninferiority of olanzapine rel- Week N Mean SD 0 207 0.78 0.18 ative to lithium was established because the 95% confi- 2 192 0.75 0.18 dence interval about the observed 8.8% absolute risk 4 183 0.77 0.19 reduction (–0.1% to 17.8%) excludes the predefined mar- 8 175 0.78 0.20 12 163 0.78 0.22 gin of noninferiority (–7.3%). Considering pole-specific re- 16 149 0.77 0.25 currences, olanzapine and lithium did not differ signifi- 20 136 0.78 0.21 cantly in the proportion of patients who had a depressive 24 120 0.77 0.25 28 116 0.76 0.23 recurrence. However, significantly fewer olanzapine- 36 101 0.79 0.24 treated patients had recurrence of manic or mixed epi- 44 87 0.73 0.27 52 75 0.79 0.23 sodes compared with lithium-treated patients. Time to symptomatic recurrence to any mood episode was not sig- (SD=4.4) for olanzapine and 1102.7 mg (SD=270.3) for lith- nificantly different between treatments (Figure 1). ium (mean serum level=0.76 meq/liter, SD=0.14). The Recurrence was further assessed as 1) meeting symp- mean lithium serum levels across the double-blind phase tomatic recurrence criteria or hospitalization for a mood are shown in Table 4. Benzodiazepines were used by sig- episode and 2) meeting DSM-IV criteria for syndromic re- nificantly more lithium-treated patients (52.3%) than currence after having met syndromic criteria for remis- olanzapine-treated patients (35.5%) (p<0.001, Fisher’s ex- sion. Rates of recurrence and odds ratios are presented in act test) during the double-blind maintenance period. An- Table 5. Considering both these criteria, significantly ticholinergics were used by 7.4% and 8.4% of olanzapine- fewer olanzapine-treated patients experienced a mood treated and lithium-treated patients, respectively (p=0.72, episode recurrence compared with lithium-treated pa- Fisher’s exact test). tients. Furthermore, as shown in Figure 1, time until mood

1284 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 TOHEN, GREIL, CALABRESE, ET AL.

TABLE 5. Mood Episode Recurrence Risk During Double-Blind Olanzapine or Lithium Maintenance Monotherapy in Bipolar Disorder Patients Following Stabilization With Olanzapine and Lithium Cotreatment Olanzapine (N=217) Lithium (N=214) Recurrence Definition and Mood Episode Type N% N%Fisher’s exact p Odds Ratio 95% CI Symptomatic recurrencea 65 30.0 83 38.8 0.055 1.5 1.0–2.2 Depression 34 15.7 23 10.7 0.15 0.6 0.4–1.1 Mania 30 13.8 50 23.4 0.02 1.9 1.1–3.1 Mixed 1 0.5 10 4.7 0.005 10.6 1.3–83.4 Symptomatic recurrencea or hospitalization 68 31.3 91 42.5 0.02 1.6 1.1–2.4 Depression 36 16.6 24 11.2 0.13 0.6 0.4–1.1 Mania 30 13.8 57 26.6 0.001 2.3 1.4–3.7 Mixed 2 0.9 10 4.7 0.02 5.3 1.1–24.3 Syndromic recurrenceb 53 26.2 69 35.8 0.05 1.6 1.0–2.4 Depression 28 13.9 16 8.3 0.11 0.6 0.3–1.1 Mania 24 11.9 49 25.4 0.001 2.5 1.5–4.3 Mixed 1 0.5 4 2.1 0.21 4.2 0.5–38.4 a Score ≥15 on Young Mania Rating Scale and/or Hamilton depression scale. b Random assignment to treatment condition was based on meeting symptomatic remission criteria; not all patients also met the DSM-IV syndromic remission criteria. Consequently, the Ns are smaller for recurrence based on DSM-IV syndromic criteria (for olanzapine, N=202; for lithium, N=193). episode recurrence was significantly longer for olanza- rum levels were not brought back into the range within the pine-treated patients. 30-day timeframe, 32 had low serum levels, and eight had Significantly fewer olanzapine-treated patients (14.3% high serum levels. Among patients with low serum levels, [N=31 of 217]) were hospitalized for a mood episode during seven (21.9%) experienced a recurrence, whereas 76 the double-blind period compared with lithium-treated (42.5%) of the 179 within the range or with high serum lev- patients (22.9% [N=49 of 214]) (p<0.03, Fisher’s exact test), els experienced a recurrence. and time to hospitalization was significantly longer for the There was no difference in recurrence rates between olanzapine group (Figure 2). For both groups, the majority lithium-treated patients with high (≥0.8 meq/liter) versus of hospitalizations was for recurrence of mania. low (<0.8 meq/liter) lithium serum levels (39.0% [N=32 of Although a documented history of intolerance or lack 82] and 39.5% [N=51 of 129], respectively; p=1.00). Fur- of response to an adequate trial of olanzapine or lithium thermore, in the 83 lithium-treated patients who experi- was an exclusion criterion, 206 such patients entered the enced a recurrence, the mean lithium serum levels were study on lithium regimens. One hundred sixty-four were slightly higher (mean=0.78 meq/liter, SD=0.11) than in the subsequently randomly assigned to double-blind treat- 128 lithium-treated patients who did not recur (mean= ment with lithium (N=80) or olanzapine (N=84). Symp- 0.75 meq/liter, SD=0.16), but the difference was not statis- tomatic recurrence criteria were met by 46.3% (N=37) of tically significant (t=1.7, df=209, p=0.09). Considering the those given lithium and 34.5% (N=29) of those given olan- 83 patients who met recurrence criteria, the mean lithium zapine (p=0.152). The impact of lithium use at study entry level before recurrence was 0.73 meq/liter (SD=0.30); how- was examined further to assess the potential bias favoring ever, for 15 of these (18.1%), the before-recurrence lithium olanzapine by comparing the differential rates of recur- level was <0.6 meq/liter. Excluding these 15 patients with rence among those who were and were not taking lithium low lithium level before recurrence provides an adjusted at entry. Among patients not taking lithium at study entry, overall rate of recurrence of 34.2% (68 of 199) for the lith- there was a 7.2% recurrence rate advantage for olanza- ium therapy group compared with 30.0% for the olanzapine (27.1%) over lithium (34.3%). Among patients taking pine group (p=0.40). The two-sided 95% confidence inter- lithium at study entry, there was an 11.8% advantage for val on this observed 4.2% risk difference (–4.8% to 13.2%) olanzapine (34.5%) over lithium (46.3%). The differential is consistent with the noninferiority of olanzapine relative advantage was not significantly different (p=0.724, Bres- to lithium. low-Day test). Additionally, the noninferiority of olanza- We further assessed whether lithium levels were a factor pine relative to lithium can be shown in each lithium-use- in study completion/disposition, using revised disposi- at-entry subgroup, since the two-sided 95% confidence tion categories in which recurrence superseded any other intervals around the risk differences did not cover the reported reason for discontinuation. The mean lithium se- predefined –7.3% margin of noninferiority. rum level for recurring lithium patients was 0.78 meq/liter (SD=0.11) and was 0.79 meq/liter (SD=0.11) for patients Lithium Levels who completed the study in sustained remission. The low- Lithium levels were obtained for 211 of 214 patients, and est mean lithium level was found among patients who dis- 171 (81%) maintained lithium levels within the therapeu- continued due to not meeting protocol criteria or noncom- tic range or were brought back into the range within the pliance (N=10, mean=0.57 meq/liter, SD=0.27). Among mandated 30-day timeframe. Of the 40 patients whose se- other reasonably sized (N>8) disposition groups, mean

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1285 OLANZAPINE VERSUS LITHIUM IN BIPOLAR MAINTENANCE

FIGURE 1. Time Until Mood Episode Recurrence Among FIGURE 2. Time Until Hospitalization Among Bipolar Disor- Bipolar Disorder Patients Randomly Assigned to Double- der Patients Randomly Assigned to Double-Blind Olanza- Blind Olanzapine or Lithium Monotherapy Following Stabi- pine or Lithium Monotherapy Following Stabilization With lization With Olanzapine and Lithium Cotreatment Olanzapine and Lithium Cotreatmenta

Symptomatic Recurrencea 1.0 1.0

0.8 0.8

0.6 0.6

0.4

0.4 0.2

0.0 Probability of Hospitalization 0.2 Lithium Symptomatic Recurrence or Hospitalizationb Olanzapine 1.0 0.0 0 50 100 150 200 250 300 350 400 0.8 Time to Hospitalization (days) a Time until hospitalization significantly longer for the olanzapine χ2 0.6 group than for the lithium group ( =6.2, df=1, p=0.01, log-rank test).

0.4 lithium levels were 0.72 meq/liter (SD=0.13) for those discontinuing due to patient decision (N=25), 0.85 meq/liter 0.2 (SD=0.12) for those discontinuing due to an adverse event (N=11), and 0.75 meq/liter (SD=0.15) for those discontinu- 0.0 ing due to lack of efficacy (N=9).

Probability of Remaining Free Recurrence To determine whether variations in lithium levels con- Syndromic Recurrencec 1.0 tributed to differences in recurrence risk at various times during the study, we examined mean lithium serum levels 0.8 for early recurrences (<150 days) and for late recurrences (>150 days). The mean serum lithium level for early recur-

0.6 rences (N=49) was 0.78 meq/liter (SD=0.11). For those patients who had a late recurrence, the mean serum level up to 150 days was 0.78 meq/liter (SD=0.11 [N=34]) and after 0.4 150 days, 0.76 meq/liter (SD=0.12 [N=32]).

0.2 Lithium Adverse Events Olanzapine One patient committed suicide during the open-label 0.0 phase of this study. During the double-blind period, two 0 50 100 150 200 250 300 350 400 patients randomly assigned to lithium died. One of these Time to Recurrence of Bipolar Disorder (days) patients committed suicide, the other died of accidental a Recurrence defined as score ≥15 on the Young Mania Rating Scale causes. and/or Hamilton depression scale. Time until recurrence longer for Thirty-four patients (6.3%) discontinued treatment dur- the olanzapine group than for the lithium group, but the difference was not significant (χ2=3.4, df=1, p=0.07, log-rank test). ing the open-label period due to an adverse event. Com- b Recurrence defined as score ≥15 on the Young Mania Rating Scale mon (≥5%) treatment-emergent adverse events during and/or Hamilton depression scale or hospitalization for a mood ep- this period were increased weight (10.3%), tremor (9.8%), isode. Time until recurrence significantly longer for the olanzapine group than for the lithium group (χ2=4.6, df=1, p=0.03, log-rank sedation (7.2%), somnolence (6.8%), and insomnia (5%). test). During the double-blind period, adverse events led to the c Recurrence defined as meeting DSM-IV criteria (other than the du- withdrawal of 41 patients in the olanzapine group (18.9%) ration criterion) for current manic, mixed, or depressive episode. Time until recurrence significantly longer for the olanzapine group and 55 patients in the lithium group (25.7%). Common or than for the lithium group (χ2=4.0, df=1, p=0.04, log-rank test). significant treatment-emergent adverse events occurring during double-blind monotherapy are reported in Table 6.

1286 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 TOHEN, GREIL, CALABRESE, ET AL.

TABLE 6. Commona or Significant Adverse Events Reported TABLE 7. Treatment-Emergent Extrapyramidal Symptoms During Double-Blind Olanzapine or Lithium Maintenance During Double-Blind Olanzapine or Lithium Maintenance Monotherapy in Bipolar Disorder Patients Following Stabi- Monotherapy in Bipolar Disorder Patients Following Stabi- lization With Olanzapine and Lithium Cotreatment lization With Olanzapine and Lithium Cotreatment Olanzapine Lithium Patients With (N=217) (N=214) Fisher’s Extrapyramidal Symptom Symptom Fisher’s Adverse Event N%N%exact p and Treatment Group N n%exact p Depression not otherwise Parkinsonism (per Simpson-Angus specified 45 20.7 25 11.7 0.01 Rating Scale) 1.00 Insomnia 17 7.8 48 22.4 <0.001 Olanzapine 177 6 3.4 Worsening of mania 17 7.8 44 20.6 <0.001 Lithium 176 5 2.8 Weight increase 14 6.5 10 4.7 0.53 Dyskinesia (per Abnormal Anxiety 12 5.5 10 4.7 0.69 Involuntary Movement Scale) Headache not otherwise At any time 0.69 specified 9 4.1 11 5.1 0.65 Olanzapine 206 3 1.5 Weight decrease 7 3.2 11 5.1 0.35 Lithium 209 2 1.0 Hypersomnia 6 2.8 0 0 0.03 At endpoint 1.00 Nausea 10.583.70.02 Olanzapine 206 0 0.0 a Occurring in ≥5% of patients in either treatment group. Lithium 209 1 0.5 Akathisia (per Barnes Rating Scale for Drug-Induced Akathisia) 0.13 We examined the occurrence of depression as a treat- Olanzapine 189 0 0.0 Lithium 197 4 2.0 ment-emergent adverse event in relation to depressive relapse among the olanzapine-treated patients. Of 45 olanzapine-treated patients with treatment-emergent de- ≥7% change from baseline. Mean change in weight during pression, 23 met relapse criteria. Most (73.9%) of these 23 the double-blind period was significantly greater for the olanzapine-treated patients had Hamilton depression olanzapine group (mean=1.8 kg, SD=5.8) than in the lith- scale total scores ≥15 at the time of the reported emer- ium group (mean=–1.4 kg, SD=5.0) (F=21.2, df=1, 385, gence of depressive symptoms, and most (65.2%) met re- p<0.001). Significantly more olanzapine-treated patients lapse criteria within 14 days of the event onset. Similarly, had ≥7% increase in weight than lithium-treated patients we examined the occurrence of insomnia as a treatment- (29.8% [N=64] versus 9.8% [N=21], respectively) (p<0.001, emergent adverse event in relation to manic symptoms Fisher’s exact test). among the lithium-treated patients. Overall, 22/48 pa- No statistically significant differences occurred between tients with insomnia met relapse criteria, but only two treatment groups in the rates of potentially clinically rele- cases of relapse were within 1 week of insomnia emer- vant changes in laboratory measures. More detailed anal- gence, and most (63%) had event onset >14 days before re- yses on nonfasting glucose and cholesterol outcomes are lapse. In addition, elevated mood (Young Mania Rating presented in Table 8. During double-blind treatment, the Scale item 1) was absent in 33 patients near the emergence mean baseline-to-endpoint change in cholesterol was of insomnia; across all 48 patients, the mean Young Mania greater for patients treated with olanzapine compared Rating Scale item 1 score was 0.49 (SD=0.86). Sleep (Young with lithium-treated patients. However, no significant dif- Mania Rating Scale item 4), however, was reduced; 24 pa- ferences occurred in incidence rates of potentially clini- tients (51%) had scores of ≥2 (sleeping less than normal by cally relevant increases in nonfasting glucose or choles- more than 1 hour) near when treatment-emergent insom- terol between treatment groups. nia was reported as an adverse event. Discussion Extrapyramidal Symptom Ratings Extrapyramidal symptoms were monitored as patient- This is the first double-blind, randomized, controlled reported treatment-emergent events (data not shown), rat- study to investigate the potential of an atypical antipsy- ing scale-defined treatment-emergent events (see Method chotic to prevent recurrence of bipolar disorder in com- section), and as mean change in scores on rating scales parison with any active treatment. Olanzapine and lith- (data not shown). Irrespective of the means of assessment, ium did not statistically differ in preventing mood episode changes in and incidences of extrapyramidal symptoms recurrence according to symptomatic rating scale criteria. were small and did not differ statistically between treat- However, olanzapine was significantly more effective than ment groups (Table 7). lithium in preventing recurrence of manic and mixed episodes. Olanzapine’s superiority to lithium in the pre- Vital Signs, Weight, and Laboratory Measures vention of mania recurrence is important because in a There were no statistically significant differences be- rigorous recently conducted meta-analysis (4), prevention tween treatments in the incidence rates of potentially clin- of mania recurrence had been identified as a particular ically relevant changes in vital signs during double-blind strength of lithium. Prevention of depression recurrence therapy. Mean weight gain during the open-label period was similar between the treatments. Time until premature was 2.74 kg (SD=3.8), and 148 (27.8%) of 532 experienced discontinuation for any reason occurred significantly ear-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1287 OLANZAPINE VERSUS LITHIUM IN BIPOLAR MAINTENANCE

TABLE 8. Nonfasting Glucose and Cholesterol Changes in Bipolar Disorder Patients Stabilized With Olanzapine and Lithium Cotreatment Then Randomly Assigned to Double-Blind Olanzapine or Lithium Monotherapy Glucose Cholesterol Open-Label Open-Label Acute Acute Cotreatment Cotreatment With With Olanzapine Double-Blind Maintenance MonotherapyOlanzapine Double-Blind Maintenance Monotherapy and Lithium Olanzapine Lithium and Lithium Olanzapine Lithium Measure (N=527) (N=207) (N=206) Analysis (N=530) (N=209) (N=206) Analysis Mean SD Mean SD Mean SD F df p Mean SD Mean SD Mean SD F df p Baseline values (mg/dl) 104.6 26.1 100.8 20.9 104.6 28.4 188.9 39.9 203.6 41.8 205.1 41.0 Baseline-to endpoint change (mg/dl) –0.4 25.6 4.7 27.2 2.3 28.3 1.99 1, 369 0.16 13.9 34.8 5.4 29.8 –12.4 29.4 24.0 1, 371 <0.001

N % N % N % p N % N % N % p <200 mg/dlb 519 98.5 206 99.5 198 96.1 298 56.2 54 25.8 55 26.2 Laboratory values– specificc 61.283.921.0 0.11a 35 11.7 4 7.4 1 1.8 0.21a a Fisher’s exact test. b At screening (open-label cotreatment) or study period entry (double-blind period). c Glucose, from <200 to ≥200 mg/dl; cholesterol, from <200 to ≥240 mg/dl lier for lithium-treated patients; the estimated median episode bipolar patients or patients with a history of a time to discontinuation was approximately 100 days rapid-cycling course were enrolled in this study, which sooner than with olanzapine. may reflect the exclusion of patient subtypes who may re- Lithium is the most extensively studied mood stabilizer; spond poorly to lithium (15). Last, serum levels of lithium meta-analyses have indicated that it is superior to placebo were comparable between those individuals who did and in the prevention of relapse (4) and reduces the risk of re- did not have a recurrence; lithium levels also did not ap- lapse 3.6-fold (5). Among placebo-controlled trials that pear to decrease at 150 days, a time when recurrence with have addressed potential bias associated with rapid with- lithium appeared to accelerate. drawal of lithium, rates of relapse with lithium were 31% Both olanzapine and lithium were generally well toler- (Bowden et al.’s 12-month study [6]), 36% (12-month out- ated. Few extrapyramidal symptom events, whether mea- comes reported by Prien et al. [13]), and 40.9% (Bowden et sured subjectively or objectively, occurred during this 52- al.’s 18-month study [7]). In these studies, lithium was su- week study. Weight gain was significantly greater in the perior to placebo in preventing bipolar disorder relapse olanzapine group. The pattern of weight gain was compara- among patients who had a manic index episode. Consis- ble with previous observations (11, 16); most of the weight tent with these reports, our results showing a recurrence gain occurred early (open-label period), followed by an ad- rate of 38.8% after 48 weeks of lithium monotherapy fur- ditional 1.8 kg gained during the maintenance period. ther support the prophylactic efficacy of lithium in bipolar No significant differences occurred between the groups disorder. in mean baseline-to-endpoint changes in nonfasting glu- Three design measures were taken to minimize bias for cose or incidences of nonfasting glucose levels ≥200 mg/ either treatment. First, patients with a history of nonre- dl. It is noteworthy, however, that this study may not have sponse, or lack of tolerance, to lithium or olanzapine were had sufficient power to determine treatment differences excluded from the study. Even though a number of patients in these adverse events. It is also noteworthy that 43.8% of were taking lithium at study entry, which may have given a patients had a nonfasting cholesterol level at screening potential selection bias advantage in favor of olanzapine in that exceeded 200 mg/dl. Measurements of the effects of preventing recurrence, the putative advantage difference atypical antipsychotics on glucose and cholesterol are im- was not significantly different (p=0.724). Second, the study portant both because of apparent increased rates of dia- design included a 4-week taper period as a means of pre- betes among patients with bipolar disorder and case re- venting recurrences associated with the abrupt withdrawal ports of diabetes among patients treated with atypical of lithium (14). Third, the enriched design of treating pa- antipsychotic agents. tients during their index episode with a combined regimen Several limitations in this study warrant discussion. of lithium and olanzapine ensured that randomly assigned There was no placebo arm in this trial, but both lithium patients were not preselected to respond preferentially to (17) and olanzapine (18) have demonstrated superior re- one treatment. Also note that small proportions of mixed- lapse prevention relative to placebo. Assessment of previ-

1288 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 TOHEN, GREIL, CALABRESE, ET AL. ous response to lithium or olanzapine was collected retro- when early remitters were excluded from the analyses, spectively and, as such, may have questionable reliability. risks of recurrence were similar to the previous values. Retrospective collection of information is a limitation of The results of this trial are consistent with previous all studies that attempt to set exclusion criteria to prevent studies demonstrating the efficacy of lithium in relapse bias associated with known lack of response to a compar- prevention in bipolar disorder and suggest that olanza- ator. This topic is an issue of all contemporary active-con- pine may also be effective in the prevention of relapse/re- trolled clinical trials and deserves more attention by clini- currence in this disorder. Additional independent studies cal trial methodologists. are needed to confirm these results. In recent meta-analyses examining the effectiveness of lithium for relapse prevention, lithium levels for six trials Acknowledgments ranged between 0.5 and 1.4 meq/liter (17). Included in The following countries and individuals participated in the these meta-analyses were two 1973 studies of Prien et al. HGHT clinical trial: Australia: Prof. P. Morris, Dr. R. Newton, Dr. D. (13), which reported median lithium levels of 0.7 and 0.8 Tannenbaum; Austria: Prof. S. Kasper, Dr. M. Schmitz, Prof. C. meq/liter; the 1971 study of Coppen et al. (19), which re- Simhandl; Belgium: Dr. G. De Bruecker, Dr. H. Bryon, Dr. B. Gillain; Bulgaria: Dr. S. Haralanov, Prof. V. Milanova, Dr. O. ported a mean lithium level of 0.93 meq/liter; the 2000 Tanchev, Prof. S. Todorov; Canada: Dr. P. Carr, Dr. L. N. Yatham; study of Bowden et al. (20), which compared valproate, Croatia: Dr. V. Jukic, Dr. N. Mandic, Prof. L. Moro; Czech Republic: lithium, and placebo and reported a mean level of lithium Dr. E. Bockova, Dr. J. Boucek, Dr. V. Hanuskova, Dr. M. Marsalek, of 1.0 meq/liter at day 30; and the 2003 study of Calabrese Dr. M. Poricky, Dr. D. Seifertova; Denmark: Dr. N. R. Hansen; Fin- land: Dr. H. J. Koponen, Dr. I. Larmo, Dr. K. Lehtinen, Dr. V. Nev- et al. (21), which compared lamotrigine, placebo, and lith- alainen, Dr. R. Riihikangas; Germany: Dr. A. Berghöfer, A/Prof. P. ium in patients with an index episode of depression and Bräunig, Prof. E. Haen, Prof. F. A. Henn, Prof. W. Maier, Prof. A. reported a mean lithium level of 0.8 meq/liter. With the ex- Marneros, Prof. H. J. Moeller, Prof. M. Schmauß; Hungary: Dr. L. ception of the Bowden et al. study (in which lithium did Haraszti, Dr. L. Mod, Dr. G. Ostorharics-Horvath, Prof. Z. Rihmer, not separate from placebo), the mean (0.76 meq/liter) and Dr. G. Vincze; Ireland: Dr. V. O’Keane, Dr. D. Walshe; Israel: Prof. H. P. Belmaker; Italy: Prof. C. Bellantuono, Prof. F. Bogetto, Prof. G. median (0.8 meq/liter) lithium levels reported in this trial B. Cassano, Prof. A. Koukopoulos, Prof. C. Maggini, Prof. G. Min- are consistent with those reported in the literature. How- nai, Prof. G.M. Muscettola; Lithuania: Dr. B. Burba, Dr. V. Maciu- ever, it is possible that some patients may not have been lis, Dr. L. E. Radavicius; the Netherlands: Prof. H. F. Kraan, Dr. J. J. maintained on optimum therapeutic levels, which may Van Egmond; New Zealand: Prof. T. Silverstone; Norway: Dr. O. represent a limitation of the study. Augland, Dr. B. Stubhaug; Poland: Dr. W. Chrzanowski, Dr. A. Czernikiewicz, Dr. J. Janczewski, Dr. J. Laczkowski, Dr. J.K. Ryba- The generalizability of this study is limited to 52 weeks kowski; Romania: Prof. P. Boisteanu, Dr. I. A. Grigoriu, Dr. A. S. Io- and mainly to individuals with a recent manic episode, nescu; Russia: Prof. L.M. Bardenstein, Prof. N. G. Neznanov, Prof. since there were few patients with a mixed index episode G. P. Panteleyeva, Prof. A. B. Smulevich; Slovak Republic: Dr. E. or a history of rapid cycling, and entry criteria excluded Palova, Dr. L. Vavrusova, Dr. L. Vircik; Slovenia: A/Prof. S. Ziherl; South Africa: Dr. C. Grobler; Sweden: Prof. R. Adolfsson, Dr. G. Ar- patients with an index episode of depression. In addition, nell, Dr. L. Haggstrom, Dr. C. Rolleri, Dr. P. Skeppar; Switzerland: the results can only be generalized to patients stabilized Dr. B. Blajev; Turkey: Prof. T. Oral; United Kingdom: Dr. C. M. with a combined regimen of olanzapine and lithium and Bonthala, Dr. A. Gregoire, Dr. D. Patience, Dr. S. Vethanayagam. may not be applicable to those stabilized with other mood stabilizers or combinations of mood stabilizers. The inter- Presented in part at the Stanley Foundation Conference on Bipolar Disorder, Freiburg, Germany, Sept. 12–14, 2002; the 41st annual pretation of recurrence on the basis of DSM-IV syndromic meeting of the American College of Neuropsychopharmacology, San criteria may be limited because the definition of recur- Juan, Puerto Rico, Dec. 8–12, 2002; the 156th annual meeting of the rence did not include a duration criterion. American Psychiatric Association, San Francisco, May 17–22, 2003; and the fifth International Conference for Bipolar Disorders, Pitts- It is noteworthy that this study may not have had suffi- burgh, June 12–14, 2003. Received Oct. 30, 2003; revisions received cient power to determine treatment differences in rare ad- June 17 and Aug. 2, 2004; accepted Aug. 11, 2004. From Lilly Re- search Laboratories; the Harvard Medical School Department of Psy- verse events and that assessment of the potential impact chiatry, McLean Hospital, Belmont, Mass.; the University of Munich of treatment on glucose homeostasis is limited in this Department of Psychiatry, Munich; the Department of Psychiatry, study because glucose and lipid measurements were non- Case Western Reserve University, Cleveland; the Harvard Medical School Department of Psychiatry, Massachusetts General Hospital, fasting. Another limitation is that more than half of the Boston; the Department of Psychiatry, University of British Columbia, patients withdrew prematurely from the study. High dis- Vancouver; Former Research Group Clinical Psychopharmacology, continuation rates in maintenance studies are common. Freie Universität Berlin, Germany; Centro Lucio Bini, Rome, Italy; the Department of Psychiatry, University of Pisa, Italy; Mood Disorders Bowden et al. (6) reported dropout rates of 62% and 76% Research Unit, Aarhus University Hospital, Aarhus, Denmark; Lilly among patients treated with divalproex and lithium for 52 Area Medical Center, Vienna, Austria; and the Department of Psychi- weeks, and dropout rates of 95%, 98%, and 100% were re- atry, University of Texas Health Sciences Center, San Antonio. Address correspondence and reprint requests to Dr. Tohen, Lilly Research ported for patients treated with lamotrigine, lithium, and Laboratories, Indianapolis, IN 46285; [email protected] (e-mail). placebo for 18 months (7). Finally, the time in remission This study was sponsored by Lilly Research Laboratories. before random assignment was relatively short; however,

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1289 OLANZAPINE VERSUS LITHIUM IN BIPOLAR MAINTENANCE

References 11. Tohen M, Ketter TA, Zarate CA, Suppes T, Frye M, Altshuler L, Zajecka J, Schuh LM, Risser RC, Brown E, Baker RW: Olanzapine 1. Coryell W, Keller M, Endicott J, Andreasen N, Clayton P, Hirsch- versus divalproex sodium for the treatment of acute mania feld R: Bipolar II illness: course and outcome over a five-year and maintenance of remission: a 47-week study. Am J Psychia- period. Psychol Med 1989; 19:129–141 try 2003; 160:1263–1271 2. Tohen M, Waternaux CM, Tsuang MT: Outcome in mania: a 4- 12. Tohen M, Baker RW, Altshuler LL, Zarate CA, Suppes T, Ketter year prospective follow-up of 75 patients utilizing survival TA, Milton DR, Risser R, Gilmore JA, Breier A, Tollefson GA: Olan- analysis. Arch Gen Psychiatry 1990; 47:1106–1111 zapine versus divalproex in the treatment of acute mania. Am 3. Tohen M, Zarate CA Jr, Hennen J, Khalsa H-MK, Strakowski SM, J Psychiatry 2002; 159:1011–1017 Gebre-Medhin P, Salvatore P, Baldessarini RJ: The McLean-Har- 13. Prien RF, Caffey EM Jr, Klett CJ: Prophylactic efficacy of lithium vard First-Episode Mania Study: prediction of recovery and first carbonate in manic-depressive illness: report of the Veterans recurrence. Am J Psychiatry 2003; 160:2099–2107 Administration and National Institute of Mental Health Collab- 4. Geddes JR, Burgess S, Hawton K, Jamison K, Goodwin GM: orative Study Group. Arch Gen Psychiatry 1973; 28:337–341 Long-term lithium therapy for bipolar disorder: systematic re- 14. Goodwin GM: Recurrence of mania after lithium withdrawal: view and meta-analysis of randomized controlled trials. Am J implications for the use of lithium in the treatment of bipolar Psychiatry 2004; 161:217–222; correction, 161:1517 affective disorder. Br J Psychiatry 1994; 164:149–152 5. Baldessarini RJ, Tondo L: Does lithium treatment still work? ev- 15. Baldessarini RJ, Tondo L, Floris G, Hennen J: Effects of rapid cy- idence of stable responses over three decades. Arch Gen Psy- cling on response to lithium maintenance treatment in 360 bi- chiatry 2000; 57:187–190 polar I and II disorder patients. J Affect Disord 2000; 61:13–22 6. Bowden CL, Calabrese JR, McElroy SL, Gyulai L, Wassef A, Petty 16. Kinon BJ, Basson BR, Gilmore JA, Tollefson GD: Long-term olan- F, Pope HG Jr, Chou JC, Keck PE Jr, Rhodes LJ, Swann AC, Hirsch- zapine treatment: weight change and weight-related health feld RM, Wozniak PJ (Divalproex Maintenance Study Group): A randomized, placebo-controlled 12-month trial of divalproex factors in schizophrenia. J Clin Psychiatry 2001; 62:92–100 and lithium in treatment of outpatients with bipolar I disorder. 17. Burgess S, Geddes J, Hawton K, Townsend E, Jamison K, Good- Arch Gen Psychiatry 2003; 57:481–489 win G: Lithium for Maintenance Treatment of Mood Disorders 7. Bowden CL, Calabrese JR, Sachs G, Yatham LN, Asghar SA, Hom- (Cochrane Review). Oxford, UK, The Cochrane Library, 2002 pland M, Montgomery P, Earl N, Smoot TM, DeVeaugh-Geiss J: 18. Tohen MF, Bowden C, Calabrese J, Chou JC, Jacobs T, Baker RW, A placebo-controlled 18-month trial of lamotrigine and lithium Williamson D, Evans AR: Olanzapine’s efficacy for relapse pre- maintenance treatment in recently manic or hypomanic pa- vention in bipolar disorder: a randomized, double-blind, pla- tients with bipolar I disorder. Arch Gen Psychiatry 2003; 60: cebo-controlled clinical trial. Eur Neuropsychopharmacol 392–400 2003; 13(suppl 4):S212 8. Hartong EG, Moleman P, Hoogduin CA, Broekman TG, Nolen 19. Coppen A, Noguera R, Bailey J, Burns BH, Swani MS, Hare EH, WA (LitCar Group): Prophylactic efficacy of lithium versus car- Gardner R, Maggs R: Prophylactic lithium in affective disorders: bamazepine in treatment-naive bipolar patients. J Clin Psychi- controlled trial. Lancet 1971; 2:275–279 atry 2003; 64:144–151 20. Bowden CL, Calabrese JR, McElroy SL, Gyulai L, Wassef A, Petty 9. Tohen M, Sanger TM, McElroy SL, Tollefson GD, Chengappa F, Pope HG Jr, Chou JC, Keck PE Jr, Rhodes LJ, Swann AC, Hirsch- KNR, Daniel DG, Petty F, Centorrino F, Wang R, Grundy SL, Gre- feld RM, Wozniak PJ (Divalproex Maintenance Study Group): A aney MG, Jacobs TG, David SR, Toma V (Olanzapine HGEH Study randomized, placebo-controlled 12-month trial of divalproex Group): Olanzapine versus placebo in the treatment of acute and lithium in treatment of outpatients with bipolar I disorder. mania. Am J Psychiatry 1999; 156:702–709 Arch Gen Psychiatry 2000; 57:481–489 10. Tohen M, Jacobs TG, Grundy SL, McElroy SL, Banov MC, Janicak 21. Calabrese JR, Bowden CL, Sachs G, Yatham LN, Behnke K, Meh- PG, Sanger T, Risser R, Zhang F, Toma V, Francis J, Tollefson GD, tonen OP, Montgomery P, Ascher J, Paska W, Earl N, DeVeaugh- Breier A: Efficacy of olanzapine in acute bipolar mania: a dou- Geiss J (Lamictal 605 Study Group): A placebo-controlled 18- ble-blind, placebo-controlled study. The Olanzapine HGGW month trial of lamotrigine and lithium maintenance treatment Study Group. Arch Gen Psychiatry 2000; 57:841-849; correc- in recently depressed patients with bipolar I disorder. J Clin tion, 2002; 59:91 Psychiatry 2003; 64:1013–1024

1290 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Service Use and Outcomes of First-Admission Patients With Psychotic Disorders in the Suffolk County Mental Health Project

Ramin Mojtabai, M.D., Ph.D., Objective: The purpose of the study was ever, the number of outpatient visits and M.P.H. to examine the inpatient and outpatient therapy sessions did not vary. Although the service use and 4-year outcomes of newly patients admitted in later years were more Daniel Herman, D.S.W., M.S. admitted psychotic patients during a pe- symptomatic at admission to their first riod of rapid change in the provision of hospitalization, their course and outcomes psychiatric services in a well-defined catch- Ezra S. Susser, M.D., Dr.P.H. over the follow-up period were not worse ment area in New York State in the 1990s. and they were not less satisfied with their Nancy Sohler, Ph.D., M.P.H. Method: Subjects were 573 participants care, compared with the patients admitted of the Suffolk County Mental Health in earlier years. Thomas J. Craig, M.D., M.P.H. Project. This group comprised patients Conclusions: The clinical characteristics with psychotic disorders first admitted of patients and the role of inpatient care Janet Lavelle, M.S. between September 1989 and August in the management of patients with psy- 1995 to 12 inpatient facilities across Suf- chotic disorders gradually changed dur- folk County, N.Y., and followed for up to 48 Evelyn J. Bromet, Ph.D. ing the 1990s. These changes, however, months. The subjects’ service use, course were not associated with changes in the of illness, symptomatic outcomes, suicide use of outpatient services or outcomes. risk, homelessness risk, and satisfaction Nevertheless, shorter hospital stays and with care were compared across admis- the presence of more severely ill patients sion years. highlight the need for more attention to Results: The length of inpatient stays de- linkage to aftercare and enhancement of creased significantly across the years. How- support networks in the community.

(Am J Psychiatry 2005; 162:1291–1298)

The 1990s witnessed many changes in mental health tween 1993 and 1995 revealed a paradoxical decrease in services for patients with severe mental disorders (1). Per- the use of outpatient services that accompanied a reduc- haps the most visible of these changes was the reduced re- tion in inpatient days (3). Such overall reduction in ser- liance on inpatient treatment. Although this trend began vices is a cause for concern, particularly for patients with in the early 1960s with the first major wave of deinstitu- severe mental disorders. Many of these patients require tionalization, it continued well through the 1990s. Be- long-term aftercare, and some require rehospitalization tween 1988 and 1994, for example, the total number of for stabilization or medication adjustment. Thus, reduced days of care in mental hospitals declined by 12.5 million use of services may adversely affect the course and out- days per year (2), a decrease that was only partly offset by come of these conditions. an increase of 1.2 million days of psychiatric care in gen- We used data from a longitudinal epidemiological study eral hospitals. This decline was due mostly to shorter lengths of hospital stays, as the number of psychiatric dis- of first-admission patients with psychotic disorders in the charges did not decrease during this period (2). early to mid-1990s in a well-defined catchment area in Policies intended to reduce the length of hospital stays New York State to examine changes in the mix and volume were partly motivated by cost concerns. Many policy mak- of services provided to this patient population during this ers and clinicians also believed that partial hospitalization period. We also examined changes in the patients’ clinical and outpatient services would be as effective as inpatient and social outcomes and in their satisfaction with care. care but would be less restrictive and more conducive to Recruitment that extended over a period of 6 years pro- patients’ integration into the community. vided a natural experiment in which systematic variations Whether and to what extent this shift in locus of care oc- in usual services and outcomes in this setting could be curred and its effects on clinical and social outcomes of studied. We focused on first admissions in order to mini- patients with severe mental illness have yet to be fully ex- mize the effect of past treatment history and better reflect amined. A study of privately insured patients treated be- the treatment careers of new entrants into the care system.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1291 FIRST-ADMISSION PSYCHOTIC PATIENTS

TABLE 1. Characteristics of Patients and Index Hospitalizations of 573 Patients Admitted in 1989–1995 to 12 Inpatient Facilities in the Suffolk County Mental Health Project, by Admission Cohort Admission Cohorts Characteristic 1989–1990 (N=104) 1990–1991 (N=101) 1991–1992 (N=74) 1992–1993 (N=71) Mean SD Mean SD Mean SD Mean SD

Age (years) 30.4 9.7 29.9 9.9 29.3 9.9 29.6 8.7

N % N % N % N % Gender Female 48 46.2 46 45.5 34 46.0 23 32.4 Male 56 53.9 55 54.5 40 54.1 48 67.6 Race/ethnicity Non-Hispanic white 79 76.0 74 73.3 58 78.4 47 66.2 Minority 25 24.0 27 26.7 16 21.6 24 33.8 Education High school or more 78 75.0 82 81.2 51 68.9 57 80.3 Did not graduate high school 26 25.0 19 18.8 23 31.1 14 19.7 Type of insurance at baseline Private 33 32.4 43 43.9 24 33.3 17 24.3 Public (Medicaid, Medicare) 17 16.7 18 18.4 10 13.9 13 18.6 Other 2 2.0 5 5.1 2 2.8 3 4.3 None 50 49.0 32 32.7 36 50.0 37 59.2 Baseline DSM-III-R research diagnosis Schizophreniab 30 28.9 26 25.7 26 35.1 22 31.0 Bipolar disorder 21 20.2 27 26.7 19 25.7 12 16.9 Major depressive disorder 15 14.4 16 15.8 14 18.9 8 11.3 Psychotic disorder not otherwise specified 5 4.8 6 5.9 7 9.5 12 16.9 Other 33 31.7 26 25.7 8 10.8 17 23.9 Lifetime alcohol/substance disorder 52 50.0 50 49.5 39 52.7 39 54.9 Type of facility Community 33 31.7 32 31.7 17 23.0 13 18.3 State 43 41.4 32 31.7 21 28.4 27 38.0 University hospital 25 24.0 31 30.7 33 44.6 28 39.4 Veterans’ hospital/other 32.965.934.134.2

Mean SD Median Mean SD Median Mean SD Median Mean SD Median

Time from onset of disorder to first admission (days) 723.5 1409.8 59.0 490.9 1232.7 21.5 572.5 1285.8 52.5 757.2 1352.7 43.5 Length of hospital stay (days) 41.0 41.0 25.5 41.5 42.3 29.0 35.6 50.8 22.5 38.9 87.8 18.0

N % N % N % N % Clinician-rated psychiatric status at discharge In full remission 54 51.9 44 44.0 21 28.4 23 32.4 In partial remission 21 20.2 19 19.0 39 52.7 37 52.1 Significant symptoms 11.055.056.834.2 Aftercare referral Day hospital 32.955.034.111.4 Outpatient clinic 75 72.1 66 66.0 44 59.5 47 66.2 Private psychiatrist 15 14.4 19 19.0 10 13.5 11 15.5 Other aftercare 7 6.7 7 7.0 8 10.8 10 14.1 Loss to follow-upc At 6-month interview 4 3.9 4 4.0 3 4.1 8 11.3 At 24-month interview 9 8.9 6 6.2 7 10.0 9 13.2 At 48-month interview 12 12.5 15 16.5 9 13.2 11 16.4 a Linear trend across the admission cohorts for categorical variables was assessed with the score test for trend of odds. (For characteristics with more than one category, data for each category were compared to data for all other categories combined.) Linear trend across the admission cohorts for continuous and ordinal variables was assessed with Spearman’s rank-order correlation (rs). b Includes schizoaffective disorder and schizophreniform disorder. c Includes participants who refused the follow-up interview or were not located. The eligible sample for follow-up comprised 552 participants at the 24-month interview and 526 participants at the 48-month interview. Based on a review of all information, including new evidence available at the 6-month consensus diagnostic meeting, 20 participants were deemed not to have been psychotic at baseline and hence were excluded from the 24-month follow-up pool. For the same reason, at the 24-month consensus meeting, an additional 19 participants were excluded from the 48-month follow-up pool. In addition, one participant had died by the 6-month follow-up, and another seven had died by the 24-month follow-up. *p<0.05. **p<0.01. ***p<0.001.

1292 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 MOJTABAI, HERMAN, SUSSER, ET AL.

We addressed three specific questions: 1) How did the use of inpatient and outpatient services change among consec- Admission Cohorts utive cohorts of patients admitted in 1989–1995? 2) How did a 1993–1994 (N=105) 1994–1995 (N=118) Analysis the 48-month course and outcome of these consecutive co- Mean SD Mean SD rs (df=571) horts of patients change? 3) How did their global function- 31.1 9.7 31.4 11.3 0.03 ing and satisfaction with care change? N % N % χ2 (df=1) Method 0.57 35 33.3 57 48.3 Subjects 70 66.7 61 51.7 0.02 The data were drawn from the Suffolk County Mental Health 79 75.2 89 75.4 Project, a longitudinal epidemiological study of consecutive first 26 24.8 29 24.6 admissions to 12 psychiatric facilities in Suffolk County, N.Y., be- 0.47 tween 1989 and 1995 (4). Briefly, inclusion criteria for the study 77 73.3 97 82.2 were age 15–60 years, residence in the county, clinical evidence of 28 26.7 21 17.8 psychosis, and both capacity and willingness to provide written 28 26.9 64 54.2 2.61 informed consent. Exclusion criteria were a psychiatric hospital- 18 17.3 13 11.0 1.04 ization more than 6 months before the current admission, mod- 3 2.9 1 1.0 0.79 erate or severe mental retardation, and inability to speak English. 55 52.9 40 33.9 0.28 Overall, 674 individuals met the inclusion criteria and agreed to participate in the study. We further limited the sample for this 31 29.5 23 19.5 1.39 study to patients who had no previous hospitalizations (N=600) 15 14.3 21 17.8 2.86 and who had their first admission between September 1, 1989, 19 18.1 23 19.5 0.86 and August 31, 1995 (N=573). Although a few patients were admitted to the participating facilities before September 1, 1989, the re- 18 17.1 23 19.5 17.92*** 22 21.0 28 23.7 1.60 cruitment sites became fully operational only after this date. Sim- 69 59.1 50 42.4 0.11 ilarly, although the study continued through the early 1996, the recruitment rate in the later months dropped to below 50%, partly 25 23.8 57 48.3 3.09 because of the extremely short stays of some of the patients. 30 28.6 23 19.5 9.61** 48 45.7 35 29.7 2.73 Data Collection 2 1.9 3 2.5 0.86 Written informed consent was obtained from the subjects for participation in the study, and their written permission to gather Mean SD Median Mean SD Median r (df=565) s information from medical records and from significant others was obtained. After the baseline interview, the subjects were in- 1058.2 1977.0 65.5 438.4 1191.2 27.5 –0.02 terviewed by telephone every 3 months and in person at months 21.7 18.2 17.5 21.0 24.2 14 –0.29*** 6, 24, and 48. Interviews were conducted by trained research interviewers, all of whom were mental health professionals. N % N % χ2 (df=1) Nonparticipation and Loss to Follow-Up The proportion of subjects who agreed to participate and were 45 42.9 54 47.8 0.20 located for interview at baseline did not vary systematically 48 45.7 37 32.7 12.34*** across the years of the study included in this report (score test for 54.8108.94.46* trend=0.10, df=1, p=0.75). Overall, 72.0% (N=674) of the patients referred to the study completed the baseline interview. Patients 10 9.5 12 10.6 7.12** 67 63.8 67 59.3 2.95 who were referred to the study but did not agree to participate or 13 12.4 25 22.1 0.55 were not located were more likely to be older and female and 7 6.7 5 4.4 0.29 more likely to have their first admission in state or university facilities rather than in community facilities or other types of facili- 10 9.5 7 5.9 2.45 ties. We adjusted for these factors in the main analyses. 15 14.6 19 16.8 6.48* At the 6-month consensus diagnostic meeting, which included 23 23.0 24 21.4 4.44* a review of all information available to the project psychiatrists, 20 participants were deemed not to have had a psychotic disorder at baseline and hence were excluded from the 24-month follow- up pool. For the same reason, at the 24-month consensus meeting, an additional 19 participants were excluded from the pool of eligible subjects for the 48-month follow-up. These exclusions had no effect on the results of the study. In addition, one participant had died by the 6-month follow-up, and another seven had died by the 24-month follow-up. Thirty-six (6.3%) of the 573 participants in the follow-up pool at 6 months, 67 (12.1%) of the 552 participants at 24 months, and 94 (17.9%) of the 526 participants at 48 months either could not be located or refused participation and were thus classified as lost to follow-up. The prevalence of loss to follow-up at 24 and 48 months increased across successive

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1293 FIRST-ADMISSION PSYCHOTIC PATIENTS

FIGURE 1. Median Lengths of First-Admission Stays for sessed by using a standardized instrument that was also com- Patients Hospitalized in State and Other Facilities in the pleted every 6 months. Suffolk County Mental Health Project, 1989–1995a Course of Illness, Clinical Outcomes, 60 Suicide, and Homelessness State facilities The course of illness was rated every 6 months by the inter- Other facilities viewers on a modified scale adopted from the Strauss-Carpenter 50 Prognostic Rating Scale (6); the period covered by the rating was the preceding 6 months. Ratings on this scale included “full remission for 3 months or longer,” “full remission for less than 3 40 months,” “partial remission,” “new episodes during interval,” and “original disorder continued.” For the analyses reported here, these categories were collapsed into the following three categories: full remission (the first two ratings), partial remission, and 30 continuous illness or new episodes (the last two ratings). Full remission was defined as an 8-week period in which the subject was

Inpatient Days asymptomatic, regardless of treatment status. Partial remission 20 was defined as having some symptoms of the index episode. Symptoms were rated at baseline and at 6, 24, and 48 months. Ratings were done with the Brief Psychiatric Rating Scale (BPRS) 10 (9), Scale for the Assessment of Positive Symptoms (SAPS) (10), and Scale for the Assessment of Negative Symptoms (SANS) (11). After each follow-up wave, a search of the National Death Index database was conducted to assess the vital status of participants 0 1989– 1990– 1991– 1992– 1993– 1994– who were lost to follow-up. On the basis of these data and further 1990 1991 1992 1993 1994 1995 information from family members of the deceased participant, (N=104) (N=101) (N=74) (N=71) (N=105) (N=118) suicidal deaths were identified. Admission Cohort Homelessness ratings were conducted at the 24- and 48-month interviews. Ratings were based on self-reports of any homeless a Dotted lines represent time trends based on fitted regression models. nights during the past 2 years. Global Functioning cohorts (Table 1). However, loss to follow-up was not associated with baseline diagnosis, severity of symptoms, course, or satisfac- Global functioning at 24 and 48 months was assessed by using tion with care at earlier assessment points. Furthermore, when a scale adopted from the SADS-L (8). Ratings included “return to analyses were repeated by limiting the sample to participants highest level of functioning,” “residual impairment,” “consider- who were not lost to follow-up by 48 months, the results were able residual impairment,” and “chronic condition or marked mainly similar to those reported here. deterioration.” Project psychiatrists made these ratings with information from interviews, medical records, interviews with sig- Patient and Index Treatment Characteristics nificant others, and rating scales. Data on patient characteristics were obtained from the Struc- Satisfaction With Care tured Clinical Interview for DSM-III-R (SCID) (5). Length of stay, The patients’ satisfaction with care was measured at 6 and 24 characteristics of the index treatment facility, patient disposition, months by using two questions: 1) How satisfied were you with and status at discharge (categorized as “full remission,” “partial the quality of services you received? (rated on a scale from 1, quite remission,” and “having significant symptoms”) were extracted dissatisfied, to 4, very satisfied) and 2) Did you get the kind of from admission hospital records by using a standardized form. treatment you wanted? (rated on a scale from 1, no, not at all, to 3, Preadmission duration of illness and treatment history were as- yes). Because some patients participated in more than one outpa- sessed as part of the SCID and were codedb by using a modified tient program, these questions were asked for as many as three version of the Strauss-Carpenter Prognostic Rating Scale (6). The programs. The ratings used here reflect the average across pro- type of insurance was determined at baseline and at the 6- and grams. There was no association between the number of pro- 24-month assessments (7). The DSM-III-R baseline research di- grams and the average rating. agnosis was a consensus decision of two or more psychiatrists on Data Analysis the basis of the SCID interviews and hospital records (4). Baseline patient characteristics were compared across six co- Course of Treatment horts identified on the basis of the date of admission (patients admitted between September 1, 1989, and August 31, 1990, formed The number and length of rehospitalizations were assessed ev- the 1989–1990 cohort, those admitted between September 1, ery 6 months. Medical records were obtained for the index admis- 1990, and August 31, 1991, formed the 1990–1991 cohort, and so sion in Suffolk County and for subsequent self-reported hospital- on). To assess trends across admission cohorts, the score test for izations and outpatient treatments, irrespective of location. On trend was used for categorical variables and Spearman’s rank- the basis of these records, the course of outpatient treatment was order correlation was used for ordinal variables. rated globally every 6 months on a scale adopted from the Sched- Patterns of treatment, illness course, and symptomatic out- ule for Affective Disorders and Schizophrenia—Lifetime Version comes over the 48 months were compared across admission co- (SADS-L) (8). Categories of response for the rating of outpatient horts by using generalized estimating equations (12). All general- treatment course included “continuous treatment,” “several brief ized estimating equation analyses were adjusted for age, gender, periods of treatment,” “consultation/brief periods,” and “none.” race, education, baseline research diagnosis, and facility type at The type and frequency of outpatient treatment contacts were as- first admission. Interaction terms for admission cohort with fol-

1294 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 MOJTABAI, HERMAN, SUSSER, ET AL. low-up time were also entered into the models. In addition, anal- received “several brief periods of treatment” (adjusted yses of service use were adjusted for updated insurance type (in- odds ratio=1.25, 95% CI=1.05–1.49, z=2.50, p<0.02). The surance type was assessed at baseline and at 6 and 24 months). In proportion of patients with “consultation/brief periods” models that showed a statistically significant linear relationship between admission cohort and outcome, we further searched for of outpatient treatment did not change systematically a possible nonlinear relationship by testing for a quadratic term across admission cohorts nor did the proportion of those for admission cohort. Generalized estimating equation analyses who received any outpatient treatment versus none. were conducted with the Stata 7 xtgee routine (13). Course of Illness, Clinical Outcomes, Results Suicide, and Homelessness Illness course and clinical outcomes for the most part Patient and Index Treatment Characteristics did not vary systematically across admission cohorts. The The patients recruited in the six admission cohorts were generalized estimating equation analyses revealed an in- similar on most sociodemographic and clinical character- crease in the proportion of patients rated as being in “full istics (Table 1). The proportion of patients recruited from remission” across admission cohorts (adjusted odds ratio= state facilities declined across the admission cohorts, 1.11, 95% CI=1.01–1.21, z=2.17, p=0.03). Further logistic mainly because of downsizing of the major adult state regression analyses revealed that the difference across co- hospital in the county. The length of first hospital stays de- horts was limited to the 6-month assessment (adjusted clined across admission cohorts. Although the reduction odds ratio=1.13, 95% CI=1.02–1.26, z=2.28, p<0.03), and in length of stay occurred in all facility types, it was partic- there were no significant differences across cohorts at the ularly dramatic for state facilities (Figure 1). later assessment points. There were also no systematic Score test for trend showed a statistically significant variations in the proportion of patients with ratings of trend in the proportion of patients discharged “in partial “partial remission” or “new episodes during interval/orig- remission” or “having significant symptoms” across inal disorder continued.” successive cohorts (Table 1). The trend for patients dis- The comparison of symptom severity in the generalized charged “in partial remission” was largely due to a dra- estimating equation analyses revealed no significant dif- matic increase in such ratings between 1991–1992 and ferences in the BPRS and SANS scores. However, analysis 1992–1993. In subsequent cohorts, however, the propor- of the SAPS scores revealed a nonsignificant trend for tion of such ratings gradually declined. There was also a higher levels of positive symptoms in later cohorts (B= systematic increase in the proportion of patients dis- 0.03, SE=0.02, z=1.89, p=0.06). Further analyses using lin- charged to day hospitals (Table 1). ear regressions revealed that the difference in SAPS scores across admission cohorts was limited to the baseline as- Course of Treatment sessment (B=0.06, SE=0.02, z=3.36, p=0.001), and there Overall, 43% of the patients were rehospitalized at least were no systematic variations at follow-up assessments. once during the 48 months of follow-up (the median num- Five (0.9%) of the 573 participants committed suicide ber of rehospitalizations among those rehospitalized was over the 48 months of the study (two in the 1990–1991 co- two, with a range from one to 12). In the generalized esti- hort and one each in the 1989–1990, 1991–1992, and 1993– mating equation analyses, the number of inpatient days 1994 cohorts). There was no systematic trend in this vari- over the 48 months declined across admission cohorts (B= able across admission cohorts (score test for trend=1.44, –1.84, SE=0.49, z=3.78, p<0.001). However, the generalized df=1, p=0.23). estimating equation analyses for the number of rehospi- At 48 months, 77 (17.1%) of the 450 participants with talizations assessed every 6 months revealed no signifi- follow-up data on housing status reported episodes of cant variations across cohorts. Thus, the decline in the homelessness since the first discharge. There was no sys- number of inpatient days was likely due to reduced tematic trend in self-reported homelessness across admis- lengths of stay, not reduced frequency of hospitalizations. sion cohorts (score test for trend=0.61, df=1, p=0.44). The reduction in inpatient days was not associated with increased use of outpatient services. In the generalized es- Global Functioning timating equation analyses, the number of day treatment, Many patients were rated as having “returned to highest individual therapy, medication, and overall outpatient vis- level of functioning” at follow-up (44.8% at 24 months and its did not systematically vary across admission cohorts. 44.1% at 48 months). In the generalized estimating equa- Furthermore, generalized estimating equation analyses of tion analyses, a significantly higher proportion of patients the global ratings of course of outpatient treatment (con- in the later admission cohorts had this rating (adjusted ducted every 6 months) suggested a decline across admis- odds ratio=1.35, 95% CI=1.13–1.63, z=3.20, p=0.001) and a sion cohorts in the proportion of patients in treatment smaller proportion of patients had a rating of “residual im- who received “continuous treatment” (adjusted odds ra- pairment” or “considerable residual impairment” (ad- tio=0.90, 95% confidence interval [CI]=0.82–0.99, z=2.23, justed odds ratio=0.79, 95% CI=0.65–0.96, z=2.33, p=0.02). p<0.03) and an increase in the proportion of patients who The proportion of patients with a rating of “chronic condi-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1295 FIRST-ADMISSION PSYCHOTIC PATIENTS tion or marked deterioration” did not vary systematically have been more symptomatic. This pattern was not very across cohorts. clear in our data, as the trend for patients discharged “in partial remission” was largely due to a dramatic increase Satisfaction With Care in such ratings between 1991–1992 and 1992–1993, and, in No differences across admission cohorts were found for subsequent cohorts, the proportion of such ratings gradu- satisfaction with care. Overall, 22.6% of the patients at 6 ally declined. However, this trend was clearly shown in an- months and 43.7% at 24 months stated that they were other study of three cohorts of depressed inpatients dis- “very satisfied” with their services. Also, 40.5% of the pa- charged between years 1988 and 1996, in which patients tients at 6 months and 64.8% at 24 months stated that they who were admitted in later years and who had shorter had received the kind of treatment that they wanted. The stays had more symptoms and a lower level of functioning increase in the proportion of satisfied patients with time after discharge (15). may be an artifact of treatment dropout of unsatisfied pa- Despite the reduced length of inpatient stays over the tients, as only the patients who received treatment during period of the study, no corresponding increase in the use the interval were asked about satisfaction with care. Fur- of outpatient services was found. Although more patients ther analyses showed that a low level of satisfaction at 6 in later admission cohorts were referred to day treatment, months was associated with a higher likelihood of drop- this pattern did not translate into increased use of day ping out of treatment during the next 6 months (satisfac- treatment services. Moreover, continuity of outpatient tion with services: odds ratio=1.65, 95% CI=1.21–2.24, z= care, rated globally, did not improve in later admission co- 3.18, p=0.001; receiving the kind of treatment that was horts. If anything, fewer patients in later cohorts received wanted: odds ratio=2.66, 95% CI=1.78–3.98, z=4.76, continuous outpatient treatment. These findings are con- p<0.001). Similarly, a low level of satisfaction at 24 months sistent with the results of a study of a national cohort of was associated with dropping out of treatment during the privately insured patients that also found no increase in next 6 months (satisfaction with services: odds ratio=2.24, utilization of outpatient services after reduction in inpa- 95% CI=1.36–3.68, z=3.16, p=0.002; receiving the kind of tient service use (3). treatment that was wanted: odds ratio=3.60, 95% CI= It is noteworthy, however, that the patients admitted dur- 1.95–6.65, z=4.09, p<0.001). ing later years experienced a speedy symptomatic recovery after discharge from their first admission and by the 6- Discussion month follow-up had symptom measures that were virtually indistinguishable from those of the patients admitted During the years of study, the mental health care system in earlier years. There was also no evidence that the course in Suffolk County, N.Y., and across the United States un- of illness in later admission cohorts was poorer than that in derwent drastic changes. One major element of these earlier cohorts. In fact, global measures showed a puzzling changes was the reduction in the length of inpatient stays. trend for patients admitted in later years to function better In New York State, these changes were expedited by a new than those admitted in earlier years. policy initiative—the Community Reinvestment Act of Although we do not have a ready explanation for these 1993—that was intended to divert funds from inpatient findings, it seems plausible that changes in the structure care to outpatient and community-based services (14). and content of services in the early to mid-1990s and, Changes in the use of inpatient services are reflected in most importantly, the drastic reduction in the length of in- our data. The average length of inpatient stays declined patient stays in this period did not adversely affect patient drastically across admission cohorts spanning the 1989– outcomes in the short run. We also did not observe any 1995 period. We also observed a systematic change in the meaningful trends in rates of suicide and homelessness characteristics of first-admission patients hospitalized in across admission cohorts. Finally, among patients who re- Suffolk County across the 6 admission years. Patients in mained in care, satisfaction with services did not vary sys- later cohorts had more severe positive symptoms at ad- tematically across cohorts. mission. This difference was probably due to changes in During the course of the study, admissions to state facil- admission policies over this period, as patients who were ities and, as a result, the proportion of state facility patients less severely ill were increasingly less likely to be admitted in this sample declined dramatically. Because patients ad- into hospitals (2). mitted to state facilities traditionally have fewer resources Both the reduction in the length of hospital stays and and experience poorer course and outcomes, the smaller the change in the characteristics of patients suggest a shift numbers of such patients in later cohorts could potentially in the role of inpatient care in the management of severely confound the results. However, when the analyses were mentally ill patients. Inpatient services were increasingly conducted separately for participants recruited from state used for short-term emergency management of more se- facilities and from other facilities, the results were essen- verely ill patients, and patients with less severe illness tially similar to the main results reported here. were shifted to less intensive settings. Probably as a result As the total number of inpatient psychiatric admissions of these changes, patients discharged in later years may in Suffolk County declined during the years of the study

1296 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 MOJTABAI, HERMAN, SUSSER, ET AL.

(16), the proportion of patients with early psychotic disor- years were more symptomatic at baseline. Second, the ders who were admitted also likely declined in later years. scope of measures of service use in this study was limited. Many of these patients may have received care in less in- Future studies need to examine other domains, including tensive settings or in criminal justice settings. However, it the process and the quality of care (17), use of informal is unlikely that sample selection could explain the findings care providers, and use of other services in the commu- of the study because the patients admitted in later years, if nity. Finally, future studies also need to examine any pos- anything, appeared to be more severely ill than those ad- sible shift of the burden of care to the criminal justice sys- mitted in earlier years. tem in this time frame. Another possible explanation for the findings of similar In conclusion, the results of this study contain a mixed course and outcome at follow-up across admission co- message for clinicians and policy makers. On the one horts despite the more severe presentation at baseline and hand, the shorter hospital stays and higher likelihood of shorter stays in later cohorts is that the potentially nega- partial remission or nonremission at the time of discharge tive effects of these factors were offset by the possible im- call for more attention to provision of community-based provements in the content of outpatient services in later support services, including reliable linkage mechanisms years, including the introduction of atypical antipsychotic that enhance continuation of aftercare in outpatient set- medications. In our sample, only 19.6% of the patients tings. Simple linking interventions (23) and focused case ever received such medications over the 48 months. Re- management programs (24) have shown promising re- peating the analyses after excluding these participants sults. The prominent role of families in the care of patients produced results similar to those reported here; thus the with severe mental disorders also calls for greater atten- findings cannot be attributed to the use of these medication to supportive and educational family interventions. tions. Nevertheless, changes in other aspects of outpatient On the other hand, it is reassuring to know that shorter care remain a possible explanation to be explored in fu- hospital stays did not negatively affect the short-term ture research. For instance, some evidence suggests a shift course and outcome of psychotic disorders and that for toward more time-limited and behavioral psychosocial in- most patients in the early course of illness, resources avail- terventions during this period at least in community men- able in the community provided adequate substitutes for tal health centers (17). hospital care. It is also plausible that the reduced length of inpatient stays had a positive effect on the course and outcome of Received April 29, 2003; revisions received Dec. 10, 2003, and March 31, 2004; accepted Aug. 9, 2004. From the Department of Psy- psychotic disorders. Past research on the relationship be- chiatry, Beth Israel Medical Center; the Departments of Psychiatry tween length of inpatient stays and clinical and social out- and Epidemiology, Mailman School of Public Health, Columbia Uni- comes produced conflicting results (15, 18–22). Perhaps versity, New York; Montefiore Medical Center and Albert Einstein Col- lege of Medicine, Bronx, N.Y.; Department of Veterans Affairs, Office most relevant to the present study are the results of the of Quality and Performance, Washington, D.C.; Hillside Hospital, Glen McLean First-Episode Psychosis Project, which recruited Oaks, N.Y.; and the Department of Psychiatry and Behavioral Sci- patients in a time frame similar to the Suffolk County Men- ence, State University of New York at Stony Brook, Stony Brook, N.Y. Address correspondence and reprint requests to Dr. Mojtabai, De- tal Health Project. That study also recorded a dramatic re- partment of Psychiatry, Beth Israel Medical Center, First Ave. at 16th duction in average length of stay during the study period St., New York, NY 10010; [email protected] (e-mail). (22). But neither time to syndromal recovery nor the pro- Supported in part by NIMH grants MH-01754 and MH-44801. portion of patients attaining syndromal recovery by 2 years varied systematically across the admission cohorts. References In interpreting the results of our study, some limitations should be considered. First, although the Suffolk County 1. Mechanic D: Mental Health and Social Policy: The Emergence Mental Health Project obtained consensus longitudinal of Managed Care. Needham Heights, Mass, Allyn & Bacon, 1999 diagnoses, we used only the baseline diagnoses in this 2. Mechanic D, McAlpine DD, Olfson M: Changing patterns of psy- report to limit the potential effect of course of illness on chiatric inpatient care in the United States, 1988–1994. Arch diagnostic decisions. It is noteworthy, however, that ad- Gen Psychiatry 1998; 55:785–791 justment of the analyses for the 24-month consensus di- 3. Leslie DL, Rosenheck R: Shifting to outpatient care? mental agnoses did not substantially change the results of the health care use and cost under private insurance. Am J Psychi- atry 1999; 156:1250–1257 study. Furthermore, the results of analyses after stratifica- 4. Bromet E, Schwartz J, Fennig S, Geller L, Jandorf L, Kovasznay B, tion of the data based on the 24-month diagnoses (schizo- Lavelle J, Miller A, Pato C, Ram R, Rich C: The epidemiology of phrenia versus other) were mainly consistent with the psychosis: the Suffolk County Mental Health Project. Schizophr results reported here. One notable difference in the sub- Bull 1992; 18:243–255 group of participants with a 24-month diagnosis of schizo- 5. Spitzer RL, Williams JBW, Gibbon M, First MB: The Structured phrenia was higher SANS scores in later cohorts. However, Clinical Interview for DSM-III-R (SCID), I: history, rationale, and description. Arch Gen Psychiatry 1992; 49:624–629 this variation was limited to the baseline assessment and 6. Strauss JS, Carpenter WT: Predictors of outcome in schizophre- did not persist at later assessment points, which suggests nia, II: relationship between predictor and outcome variables. that the participants with schizophrenia admitted in later Arch Gen Psychiatry 1974; 31:37–42

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1297 FIRST-ADMISSION PSYCHOTIC PATIENTS

7. Rabinowitz J, Bromet EJ, Lavelle J, Severance KJ, Zariello SL, 17. Cypres A, Landsberg G, Spellman M: The impact of managed Rosen B: Relationship between type of insurance and care dur- care on community mental health outpatient services in New ing the early course of psychosis. Am J Psychiatry 1998; 155: York State. Adm Policy Ment Health 1997; 24:509–521 1392–1397 18. Riessman CK, Rabkin JG, Struening EL: Brief versus standard 8. Endicott J, Spitzer RL: A diagnostic interview: the Schedule for psychiatric hospitalization: a critical review of the literature. Affective Disorders and Schizophrenia. Arch Gen Psychiatry Community Ment Health Rev 1977; 2:1–10 1978; 35:837–844 19. Heeren O, Dixon L, Gavirneni S, Regenold WT: The association 9. Woerner MG, Mennuzza S, Kane JM: Anchoring the BPRS: an between decreasing length of stay and readmission rate on a aid to improved reliability. Psychopharmacol Bull 1988; 24: psychogeriatric unit. Psychiatr Serv 2002; 53:76–79 112–117 20. Craig TJ, Fennig S, Tanenberg-Karant M, Bromet EJ: Rapid ver- 10. Andreasen NC: Scale for the Assessment of Positive Symptoms sus delayed readmission in first-admission psychosis: quality (SAPS). Iowa City, University of Iowa, 1984 indicators for managed care? Ann Clin Psychiatry 2000; 12: 11. Andreasen NC: Scale for the Assessment of Negative Symptoms 233–238 (SANS). Iowa City, University of Iowa, 1983 21. Mojtabai R, Nicholson RA, Neesmith DH: Factors affecting re- 12. Carr GJ, Chi EM: Analysis of variance for repeated measures: lapse in patients discharged from a public hospital: results data: a generalized estimating equations approach. Stat Med from survival analysis. Psychiatr Q 1997; 68:117–129 1992; 11:1033–1040 13. Stata Reference Manual: Release 7.0. College Station, Tex, Stata 22. Tohen M, Hennen J, Zarate CM Jr, Baldessarini RJ, Strakowski Corp, 2001 SM, Stoll AL, Faedda GL, Suppes T, Gebre-Medhin P, Cohen BM: 14. New York State Office of Mental Health: Reinvestment’s First Two-year syndromal and functional recovery in 219 cases of Five Years, 2000. http://www.omh.state.ny.us/omhweb/omhq/ first-episode major affective disorder with psychotic features. q1299/reinvestment%5Freview.htm Am J Psychiatry 2000; 157:220–228 15. Lieberman PB, Wiitala SA, Elliott B, McCormick S, Goyette SB: 23. Olfson M, Mechanic D, Boyer CA, Hansell S: Linking inpatients Decreasing length of stay: are there effects on outcomes of with schizophrenia to outpatient care. Psychiatr Serv 1998; 49: psychiatric hospitalization? Am J Psychiatry 1998; 155:905–909 911–917 16. Mental Health Statistics Unit, New York State Office of Mental 24. Susser E, Valencia E, Conover S, Felix A, Tsai WY, Wyatt RJ: Pre- Health: 2000 New York State Chartbook of Mental Health In- venting recurrent homelessness among mentally ill men: a formation. http://www.omh.state.ny.us/omhweb/chartbook/ “critical time” intervention after discharge from a shelter. Am J text.htm Public Health 1997; 87:256–262

1298 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Pathways to PTSD, Part I: Children With Burns

Glenn N. Saxe, M.D., F.R.C.P. Objective: The goal of this study was to Results: Two pathways to PTSD were dis- develop a model of risk factors for post- cerned: 1) from the size of the burn and Frederick Stoddard, M.D. traumatic stress disorder (PTSD) in a level of pain following the burn to the group of acutely burned children. child’s level of acute separation anxiety, Erin Hall, M.A. Method: Seventy-two children between and then to PTSD, and 2) from the size of the burn to the child’s level of acute disso- Neharika Chawla, M.A. the ages of 7 and 17 who were admitted to the hospital for an acute burn were el- ciation following the burn, and then to PTSD. Together these pathways ac- Carlos Lopez, M.D. igible for study. Members of families who consented completed the Child PTSD Re- counted for almost 60% of the variance in Robert Sheridan, M.D. action Index, the Multidimensional Anxi- PTSD symptoms and constituted a model ety Scale for Children, and other self-re- with excellent fit indices. Daniel King, Ph.D. port measures of psychopathology and Conclusions: These findings support a environmental stress both during the hos- model of complex etiology for childhood Lynda King, Ph.D. pitalization and 3 months following the PTSD in which two independent path- burn. A path analytic strategy was used to ways may be mediated by different Rachel Yehuda, Ph.D. build a model of risk factors for PTSD. biobehavioral systems.

(Am J Psychiatry 2005; 162:1299–1304)

A pproximately 1.25 million individuals received burn order. Acute stress disorder is the psychopathological re- injuries in the United States in 1992 (1). Fire is the third sponse in the immediate aftermath of a traumatic event leading cause of unintentional injuries in children (2). Not that occurs until 1 month following the trauma and in- surprisingly, burns were among the first types of trauma cludes both anxiety and dissociative symptoms as neces- identified as leading to traumatic stress reactions in sary components of this diagnosis. Daviss and colleagues adults, and early burn research contributed to the original (11) and Koplin Winston and colleagues (12) reported a classification of posttraumatic stress disorder (PTSD) in broad range of acute stress disorder symptoms in separate DSM-III (3–5). Early studies of burned children alerted re- cohorts of injured children. Numerous studies have docu- searchers and clinicians to the importance of attending to mented the higher prevalence of PTSD in those initially di- their psychiatric needs (6, 7). In one of the first studies of agnosed with acute stress disorder (13–15). the prevalence of psychiatric diagnoses, 30% of the chil- There is, however, considerable controversy over which dren interviewed more than 6 months after a severe burn symptom cluster, anxiety or dissociation, is more predic- had met the DSM-III criteria for PTSD at some point after tive of PTSD and, regarding the diagnosis of acute stress their burn. These authors also reported higher preva- disorder, whether dissociation should be included as a lences of overanxious disorder, phobias, and enuresis in distinct symptom cluster at all. If dissociative symptoms the children with burns than in a nonburn comparison group (8, 9). However, as little is known about the etiology are required in the diagnosis of acute stress disorder, many of PTSD in children with burns, there is a great need for individuals who suffer from only anxiety symptoms may risk factor studies. Further, as burned children are accessi- not receive adequate attention (16, 17). Further, it has ble for study shortly after their trauma and, by virtue of been argued that dissociative symptoms are not necessar- their ongoing need for medical and surgical care, can be ily more predictive of PTSD than are anxiety symptoms readily followed over time, they are ideally suited to pro- (16). On the other hand, evidence suggests that individu- spective, longitudinal research designs. Such designs are als who dissociate around the time of trauma are at high necessary for the advancement of the understanding of risk of developing PTSD (18–21). The controversy over the risk factors for PTSD (10). In this study we attempted to relative importance of anxiety and dissociative symptoms build a model of risk factors for PTSD symptoms in burned is also important for another reason: these two groupings children by assessing them shortly after the burn and then of symptoms may be the phenotypes of different biobe- 3 months following this assessment. havioral systems related to PTSD (22, 23). Clarity regarding Research has focused on the anxiety/arousal and the the relationship between these groups of symptoms and dissociative symptoms expressed in the acute aftermath of their biological underpinnings may be critical for identify- a trauma. The importance of these symptom clusters is ing children at risk and refining immediate interventions formalized in DSM-IV with the addition of acute stress dis- following a trauma.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1299 PTSD IN CHILDREN WITH BURNS

TABLE 1. Pretrauma, Trauma, Peritrauma, and Posttrauma questions about the child’s dissociative symptoms (interview and Variables for 72 Children With an Acute Burn questionnaires are described in the following). Variable Mean SD Follow-up assessments consisting of the same interviews and Pretrauma: age (years) 11.20 3.51 questionnaires were completed by a trained research associate Trauma: total body surface area burned (%) 17.58 17.92 (N.C. or E.H.) at the participant’s home 3 months later. Peritrauma Separation anxiety (score on Multidimensional Measures Anxiety Scale for Children [25], possible Child measures. The Child PTSD Reaction Index (24) is a 20- range=0–27) 8.21 5.25 item semistructured interview that assesses posttraumatic symp- Dissociation (score on Child Stress Disorders toms in children. Its interrater reliability is high (Cohen’s kappa= Checklist [27], possible range=0–16) 3.19 2.90 0.88). Its validity is supported by the finding that children who are Pain (score on Colored Analogue Pain Scale [26], possible range=0–10) 2.12 2.24 known to have PTSD have much higher scores on this instrument Posttrauma: PTSD symptoms 3 months after (24). The score on the Child PTSD Reaction Index was the main burn (score on Child PTSD Reaction Index [24], dependent variable in this study. possible range=0–80) 16.82 13.13 The Multidimensional Anxiety Scale for Children (25) is a 39- item self-report measure of pediatric anxiety symptoms. In a psychometric study of the scale, the mean intraclass correlation coef- Of the anxiety symptoms, we regard separation anxiety ficients at 3 weeks and 3 months were 0.79 and 0.93, respectively, as particularly important for the population of children demonstrating satisfactory to excellent test-retest reliability (25). hospitalized with burns. The nature of burn injury and its Factor analytic studies of the scale have shown a variety of inde- long and stressful hospital course frequently mean separa- pendent scales, including “harm avoidance,” “separation anxiety,” and “physical symptoms.” The separation anxiety scale of tions of the child and the parents at a time when the child the Multidimensional Anxiety Scale for Children was our index of has a great need for their help and comfort. separation anxiety. In the current study we used path analytic techniques to The Colored Analogue Pain Scale (26) is a pocket-sized visual evaluate the ways in which pretrauma variables, trauma analogue instrument on which the child slides a marker along a 10-cm line that shows an increasing intensity of red color corre- characteristics, and reactions in the immediate aftermath sponding to increased intensity of current pain. This instrument of a trauma are related to later PTSD symptoms. As de- has been used with many groups of children who have pain. Chil- scribed, anxiety and dissociative reactions in the wake of a dren with more painful syndromes score higher on this instru- trauma have frequently been identified as important comment than do children with less painful syndromes. The Colored ponents of the acute trauma response, although there is Analogue Pain Scale has been found to be easier to administer than other visual analogue scales. considerable controversy over which is more strongly related to deleterious long-term outcomes. Consequently, a Nurse measures. The numbing and dissociation scale of the Child Stress Disorders Checklist (27) is a measure of acute disso- main goal of the current study was to assess the relative ciative symptoms based on observer report. This 8-item scale as- importance of anxiety and dissociation in the immediate sesses the dissociative dimension of the child’s acute and post- aftermath of a burn. traumatic stress response. Internal consistency was found to be 0.75 (Cronbach’s alpha). Test-retest reliability, calculated by cor- relating scores reported by the parents 2 days apart, was found to Method be 0.72 (intraclass correlation). In the current study, the child’s Participants primary nurse completed the Child Stress Disorders Checklist in regard to the child’s dissociative symptoms. The participants were drawn from a group of children admitted The total body surface area burned was the percentage re- to Shriners Burns Hospital in Boston for an acute burn. All chil- corded by the attending surgeon in the child’s medical record. dren ages 7 to 17 years were eligible to participate unless they or their parents did not speak sufficient English to complete the Data Analysis study instruments. Of 116 eligible children, 72 (62%) participated. We used a path analytic strategy similar to that used by Shalev Of the 44 children who did not, 26 had families that declined to et al. (20) in another prospective study of acutely traumatized in- participate and 18 were discharged before we were able to obtain dividuals. As this strategy is based on a prospective longitudinal consent. The mean age of the participants was 11.20 years (SD= method, the directionality of many of the paths was constrained 3.51); 24 were girls and 48 were boys. The average length of stay by the time at which the variables were assessed. Accordingly, we was 25 days (SD=23). The mean amount of body surface area divided variables into the following: 1) PTSD symptoms (our burned was 17.58% (range=1%–85%). The children were inter- main dependent variable, derived from the Child PTSD Reaction viewed an average of 10 days after admission (range=2–26 days). Index), 2) posttraumatic variables (variables assessed at the 3- month follow-up), 3) peritraumatic variables (variables assessed Procedures shortly after the trauma), 4) trauma exposure variable (percent- Within 3 days of hospitalization or when the child was consid- age of body surface area burned), and 5) pretrauma variables ered medically stable (e.g., did not have a delirium, did not have (variables related to the child or family from before the trauma). an active infection, and was not receiving mechanical ventila- A series of hierarchically nested ordinary least squares multiple tion), the child and his or her parents were approached by one of regression analyses were used to estimate direct and indirect the investigators (C.L.) and introduced to the study. After com- effects among variables. The first step was to predict the depen- plete description of the study, written informed consent and as- dent variable (PTSD symptoms). From the remaining variables, sent were obtained from the parents and child, respectively. On we chose combinations of variables that accounted for a high the same day that consent and assent were obtained, the child percentage of the variance in PTSD symptoms (high R2 value), was interviewed and answered questions about his or her trau- guided by our theoretical model of PTSD and constrained by the matic stress responses and the child’s primary nurse answered strength of bivariate relationships (all bivariate relationships with

1300 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 SAXE, STODDARD, HALL, ET AL.

TABLE 2. Correlations Among Pretrauma, Trauma, Peritrauma, and Posttrauma Variables for 72 Children With an Acute Burn Correlation (r) Variable Total Body Surface Area Burned Separation Anxiety Dissociation Pain PTSD Symptoms Pretrauma: age –0.12 –0.42a** –0.03 0.02 –0.17 Trauma: total body surface area burned 0.51a** 0.31a* 0.17 0.49** Peritrauma Separation anxiety 0.13 0.52a** 0.68a** Dissociation 0.06 0.36a* Pain 0.45** Posttrauma: PTSD symptoms a A direct relationship exists between these variables. *<0.01. **<0.001. r>0.30 and p<0.01). Accordingly, we chose two primary variables FIGURE 1. Path Analytic Model for the Development of (i.e., separation anxiety and dissociative symptoms measured in PTSD in 72 Children With an Acute Burna the hospital) that together accounted for 59% of the variance of PTSD (R2=0.59). Once these two variables were chosen, antecedent variables could be identified in order to place separation anx- –0.33 Separation Age iety and dissociation in the roles of mediators. Again, we were anxiety guided by theory and the strength of the bivariate relationships, 0.42 0.68 and we were further constrained by temporal relationships. The 0.50 direction of the relationships chosen had to make temporal sense Body (e.g., peritraumatic dissociation could not lead to the size of the surface PTSD Pain burn). In this way we generated a network of associations among area variables that, as we will discuss, adds important understanding burned to the unfolding of PTSD over time. 0.31 0.26 Problems related to missing data can reduce the number of subjects for a particular analysis to a less than optimal level. We Dissociation employed the statistical package M Plus 2.1 (28) with full information maximum likelihood estimation to retain the full number of subjects for each analysis. M Plus is preferred since it is able to use the full information maximum likelihood procedure in concert Pretrauma Trauma Peritrauma Posttrauma with the Satorra-Bentler correction for nonnormal data. (See the a work of McArdle [29] and Graham et al. [30] for discussions of the The values in the model are partial correlation coefficients (beta 2 advantages of maximum-likelihood-based methods for incom- weights). R =0.59. plete data over more traditional listwise and pairwise deletion procedures.) dissociative responses. The magnitude of the trauma, measured by the size of the burn, was not related to PTSD di- Results rectly but exerted its influence indirectly through both Table 1 presents the mean values and standard devia- pathways. The pathway mediated by separation anxiety tions for the variables used in our path analysis, and Table was influenced by the acute pain response. This pathway 2 presents a correlation matrix of these variables. was also influenced by the size of the burn and was inversely related to the age of the child. The pathway medi- Figure 1 illustrates the network of associations among ated by the acute dissociative response was influenced only variables. The results of the path analysis indicate two di- by the size of the burn. rect pathways to PTSD. These pathways are from 1) acute separation anxiety (beta=0.68) and 2) acute dissociation The independence of the anxiety and dissociation path- (beta=0.26). In addition, separation anxiety served as a ways to PTSD suggests the possibility that different biobe- mediator between the age of the child and PTSD and behavioral systems contribute to PTSD. A number of re- tween pain measured shortly after the burn and PTSD. searchers have drawn connections between the arousal/ Moreover, both separation anxiety and dissociation medi- anxiety symptoms and the sympathetically mediated ated the relationship between total burn area and PTSD. fight-or-flight response and between the symptoms of dis- Together these pathways account for 59% of the variance sociation and the parasympathetically mediated “freeze” in PTSD. The overall model yielded strong fit indices (χ2= or “immobilization” response (22, 23, 31, 32). For example, 2.0, df=6, p=0.92; comparative fit index=1.00; Tucker- Perry and colleagues have described the hierarchical re- Lewis index=1.00; and root mean square error of approxi- sponse to threat in biological theories of children’s re- mation=0.00). sponses to trauma (22, 23). They described the initial fight-or-flight hyperarousal response in a child faced with Discussion an immediate threat and the following freeze-or-surrender immobilized response when the child cannot dimin- Two pathways to PTSD symptoms were found. One path- ish the threat by means of the fight-or-flight response. The way was mediated by separation anxiety, the other by acute authors suggested that this freeze-or-surrender response

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1301 PTSD IN CHILDREN WITH BURNS occurs when a child is confronted by extreme threat and is hypotheses of Perry and others that the anxiety/arousal helpless to respond. This response is phenotypically ob- component of PTSD may be a phenotype of the sympa- served as dissociative symptoms (22). The fight-or-flight thetically mediated fight-or-flight response, whereas the response is controlled by the sympathetic/HPA axis sys- dissociative symptoms may be a phenotype of the para- tem, and the freeze/immobilization response is controlled sympathetically mediated immobilization or freezing re- by the parasympathetic nervous system (33). Porges has sponse. This response is described as phylogenetically described the evolutionary foundations of this sequen- very old and is characterized by the lack of vagal tone, tially overwhelmed autonomic nervous system (33). Fur- bradycardia, and shutting down of responses, in order to ther, Bowlby (34, 35) has noted the critical interpersonal conserve resources to maximize the chance of survival components of this threat response system in humans and during situations of extreme threat. This type of response nonhuman primates. is thought to occur after exhaustion of other defensive behaviors, such as the fight-or-flight response, which is me- Anxiety/Arousal Pathway to PTSD diated by the sympathetic nervous system (33, 38, 39). Bowlby stated, “Of the many fear arousing situations that Changes in vagal tone, a well-accepted marker of para- a child, or older person, can foresee, none is likely to be sympathetic activity, have been associated with PTSD (40– more frightening than the possibility that an attachment 42). It may be that situations of extreme life threat lead to figure will be absent or…unavailable when needed” (35, p. the parasympathetically mediated shutting down of emo- 201). In this study, the child’s level of acute separation anx- tional responses, phenotypically observed as dissociative iety was directly related to PTSD symptoms. Separation symptoms and prospectively related to PTSD. Parasympa- anxiety was influenced by the child’s age, the size of the thetic nervous system activation was not, however, di- burn, and the degree of pain experienced. As described, rectly measured in this study. the nature of burn injury and the prolonged recovery pro- It is notable that no direct relationship between separa- cess during hospitalization mandate stressful separations tion anxiety and dissociation was found in our model. In between the child and parents at a time when the child has fact, the bivariate association (Table 2) was negligible (r= a great need for their help and comfort. Thus, the children 0.13). As dissociative responses are hypothesized to follow who experienced the most anxiety on separations were anxiety/arousal responses when the fight-or-flight ap- more likely to develop posttraumatic symptoms. Separa- proach is ineffective, it would follow that anxiety would tion anxiety in children may be related to the evolutionarily lead to dissociation. This dependence of dissociation on driven reaction to the anticipated loss of the mother and of anxiety/arousal, but not vice versa, is reported in the ac- her protective function (36). As survival is a frequent con- companying article in this issue of the Journal (43). It is cern of children on burn units, the need for soothing and possible that variables that were not assessed (such as reassurance from a parent is felt intensely by most burned psychophysiology) may be mediating this relationship. children. This extreme arousal response to being burned, Regarding the controversy of whether acute anxiety or in pain, and alone is consistent with our data. dissociation is the more important predictor of PTSD, our It is noteworthy that burn trauma has been considered a data suggest that both anxiety and dissociative symptoms relatively impersonal trauma, as it is often not caused by independently contribute to the risk for PTSD. another person. The importance of separation anxiety in the development of PTSD suggests that burn trauma has a Limitations significant interpersonal component. Specifically, burn This study is limited by a relatively small number of sub- trauma and the emotional distress underlying the ques- jects and a short longitudinal follow-up. The findings of tion “Who will help me?” are integrally connected. This ev- two independent pathways to PTSD, and our interpreta- idence thereby supports the notion that all trauma has an tion that they suggest discrete biobehavioral systems, can interpersonal element. be illuminated by studies that include psychophysiological and neuroendocrine measures. Dissociative Pathway to PTSD Krystal has stated, “The switch from anxiety to the cata- Clinical Implications tonoid response is the subjective evaluation of impending These data have important implications for PTSD treat- danger as one that cannot be avoided or modified. With ment and prevention. If symptoms of separation anxiety, the perception of fatal helplessness in the face of destruc- pain, and dissociation can be identified in the acute after- tive danger, one surrenders to it” (37, pp. 114–115). In this math of a trauma and strongly contribute to the risk of study, the degree of dissociative symptoms measured PTSD, then it is critical to assess these symptoms and to shortly after the burn was found to be a direct predictor of intervene accordingly. There are well-described psychoso- PTSD symptoms. This replicates the results of many stud- cial and biological interventions for anxiety, pain, and dis- ies (18, 19), but we believe that it is the first to document sociation. Given the particular importance of separation this effect in children with burns. The independence of anxiety, burn hospitals (and all pediatric intensive care dissociation from separation anxiety is consistent with the units) must make every effort to keep parents and children

1302 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 SAXE, STODDARD, HALL, ET AL. together and to work with parents to increase their capac- stress disorder following mild traumatic brain injury. Am J Psy- ity to comfort their children. It is noteworthy that opiates chiatry 2000; 157:626–628 are among the strongest inhibitors of the distress cry in 14. Bryant RA, Harvey AG, Guthrie R, Moulds M: A prospective study of acute psychophysiological arousal, acute stress disor- young animals upon separation from mothers (35, 44). We der, and posttraumatic stress disorder. J Abnorm Psychol 2000; have previously reported that the dose of morphine re- 109:341–344 ceived by burned children in the hospital diminishes 15. Holeva V, Tarrier N, Wells A: Prevalence and predictors of acute PTSD symptoms over time (45). It is possible that one stress disorder and PTSD following road traffic accidents: mechanism for the diminution of PTSD over time in these thought control strategies and social support. Behav Ther 2001; 32:65–83 children is the effect of the morphine on separation anxi- 16. Bryant RA, Harvey AG: Acute Stress Disorder: A Handbook of ety, as well as pain. If so, it would be a very specific inter- Theory, Assessment, and Treatment. Washington, DC, Ameri- vention for one of our two pathways to PTSD. can Psychological Association, 2000 17. Marshall RD, Spitzer R, Liebowitz MR: Review and critique of Presented at the 18th annual meeting of the International Society the new DSM-IV diagnosis of acute stress disorder. Am J Psychi- for Traumatic Stress Studies, Baltimore, Nov. 7–10, 2002. Received atry 1999; 156:1677–1685 Dec. 12, 2003; revision received July 20, 2004; accepted Aug. 2, 18. Marmar CR, Weiss DS, Schlenger WE, Fairbank JA, Jordan BK, 2004. From the Department of Child and Adolescent Psychiatry, Bos- Kulka RA, Hough RL: Peritraumatic dissociation and posttrau- ton University School of Medicine; the Department of Psychiatry, matic stress in male Vietnam theater veterans. Am J Psychiatry Shriners Burns Hospital, Boston; and the Department of Psychiatry, 1994; 151:902–907 Bronx VA Medical Center, Bronx, N.Y. Address correspondence and reprint requests to Dr. Saxe, Department of Child and Adolescent Psy- 19. Koopman C, Classen C, Spiegel D: Predictors of posttraumatic chiatry, Boston University School of Medicine, Dowling 1 North, 1 stress symptoms among survivors of the Oakland/Berkeley, Ca- Boston Medical Center Place, Boston, MA 02118; glenn.saxe@ lif, firestorm. Am J Psychiatry 1994; 151:888–894 bmc.org (e-mail). 20. Shalev AY, Peri T, Canetti L, Schreiber S: Predictors of PTSD in in- Supported by NIMH grant R01 MH-57370 and by Substance Abuse jured trauma survivors: a prospective study. Am J Psychiatry and Mental Health Services Administration grant U79 SM-54305 (Dr. 1996; 153:219–225 Saxe). 21. Difide JA, Barocas DA, Eskra D, Jaffe A, Roberts J, Yurt RW: Acute stress disorder following burn injury, in 1997 Annual Meeting Syllabus and Proceedings Summary. Washington, DC, Ameri- References can Psychiatric Association, 1997, symposium 72 1. Brigham PA, McLoughlin E: Burn incidence and medical care 22. Perry BD, Pollard RA, Blakley TL, Baker WL, Vigilante D: Child- use in the United States: estimates, trends, and data sources. J hood trauma, the neurobiology of adaptation, and use-depen- Burn Care Rehabil 1996; 17:95–107 dent development of the brain: how states become traits. In- 2. Health and the American Child: Risks, Trends, and Priorities for fant Ment Health J 1995; 16:271–291 the 21st Century. Washington, DC, Public Health Policy Advi- 23. Perry BD: Incubated in terror: neurodevelopmental factors in sory Board, 1999, p 25 the “cycle of violence,” in Children in a Violent Society. Edited 3. Andreasen NJ: Neuropsychiatric complications in burn pa- by Osofsky J. New York, Guilford, 1997, pp 124–148 tients. Int J Psychiatry Med 1974; 5:161–171 24. Nader KO: Assessing trauma in children: self-reports of DSM 4. Andreasen NJ, Noyes R Jr, Hartford CE: Factors influencing ad- Criteria B–D symptoms, in Assessing Psychological Trauma and justment of burn patients during hospitalization. Psychosom PTSD. Edited by Wilson JP, Keane TM. New York, Guilford, 1996, Med 1972; 34:517–525 pp 513–537 5. Andreasen NJ, Hartford CE, Knott JR, Canter A: EEG changes as- 25. March JS, Parker JDA, Sullivan K, Stallings P, Conners K: The sociated with burn delirium. Dis Nerv Syst 1977; 38:27–31 Multidimensional Anxiety Scale for Children (MASC): factor 6. Seligman R: A psychiatric classification system for burned chil- structure, reliability, and validity. J Am Acad Child Adolesc Psy- dren. Am J Psychiatry 1974; 131:41–46 chiatry 1997; 36:554–565 7. Seligman R, Macmillan BG, Carroll SS: The burned child: a ne- 26. McGrath PA, Brigham MC: The assessment of pain in children glected area of psychiatry. Am J Psychiatry 1971; 128:52–57 and adolescents, in Handbook of Pain Assessment. Edited by 8. Stoddard FJ: Care of infants, children, and adolescents with Turk DC, Melzack R. New York, Guilford, 1992, pp 295–314 burn injuries, in Child and Adolescent Psychiatry: A Compre- 27. Saxe G, Chawla N, Stoddard F, Kassam-Adams N, Courtney D, hensive Textbook. Edited by Lewis M. Baltimore, Williams & Cunningham K, Lopez C, Hall E, Sheridan R, King D, King L: Wilkins, 1995, pp 1016–1037 Child Stress Disorders Checklist: a measure of ASD and PTSD in 9. Stoddard FJ, Norman DK, Murphy JM, Beardslee WR: Psychiat- Children. J Am Acad Child Adolesc Psychiatry 2003; 42:972–978 ric outcome of burned children. J Am Acad Child Adolesc Psy- 28. Muthen LK, Muthen BO: M Plus Statistical Analysis With Latent chiatry 1989; 28:589–595 Variables: User’s Guide. Los Angeles, Muthen & Muthen, 1998 10. Yehuda R: Risk Factors for Posttraumatic Stress Disorder. Wash- 29. McArdle JJ: Structural factor analysis with incomplete data. ington, DC, American Psychiatric Press, 1999 Multivariate Behav Res 994; 29:409–454 11. Daviss WB, Racusin R, Fleischer A, Mooney D, Ford JD, McHugo 30. Graham JW, Hofer SM, Donaldson SI, MacKinnon DP, Schafer JL: G: Acute stress disorder symptomology during hospitalization Analysis with missing data in prevention research, in The Sci- for pediatric injury. J Am Acad Child Adolesc Psychiatry 2000; ence of Prevention: Methodological Advances From Alcohol 39:569–575 and Substance Abuse Research. Edited by Bryant KJ, Windle M, 12. Winston FK, Kassam-Adams N, Vivarelli-O’Neill C, Ford J, New- West SG. Washington, DC, American Psychological Association, man E, Baxt C, Stafford P, Cnaan A: Acute stress disorder symp- 1997, pp 325–366 toms in children and their parents after pediatric traffic injury. 31. Perry BD: The memories of states: how the brain stores and re- Pediatrics 2002; 109:e90 trieves traumatic experience, in Splintered Reflections: Images 13. Harvey AG, Bryant RA: Two-year prospective evaluation of the of the Body in Trauma. Edited by Goodwin JM, Attias R. New relationship between acute stress disorder and posttraumatic York, Basic Books, 1999, pp 9–38

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1303 PTSD IN CHILDREN WITH BURNS

32. Scaer R: The neurophysiology of dissociation and chronic dis- 40. Sahar T, Shalev AY, Porges SW: Vagal modulation of responses ease. Appl Psychophysiol Biofeedback 2001; 26:73–91 to mental challenge in posttraumatic stress disorder. Biol Psy- 33. Porges SW: Orienting in a defensive world: mammalian modi- chiatry 2001; 49:637–643 fications of our evolutionary heritage: a polyvagal theory. Psy- 41. Cohen H, Kotler M, Matar MA, Kaplan Z, Loewenthal U, Mi- chophysiology 1995; 32:301–318 odownik H, Cassuto Y: Analysis of heart rate variability in post- 34. Bowlby J: Attachment and Loss, vol I: Attachment. New York, traumatic stress disorder patients. Biol Psychiatry 1998; 44: Basic Books, 1969 1054–1059 35. Bowlby J: Attachment and Loss, vol II: Separation: Anxiety and 42. Cohen H, Kotler M, Matar MA, Kaplan Z, Miodownik H, Cassuto Anger. New York, Basic Books, 1973 Y: Power spectral analysis of heart rate variability in posttrau- 36. Panksepp J: Loneliness and the social bond: the brain sources matic stress disorder patients. Biol Psychiatry 1997; 41:627– of sorrow and grief, in Affective Neuroscience: The Founda- 629 tions of Human and Animal Emotions. Edited by Panksepp J. 43. Kaplow JB, Dodge KA, Amaya-Jackson L, Saxe GN: Pathways to New York, Oxford University Press, 1998, pp 261–279 PTSD, part II: sexually abused children. Am J Psychiatry 2005; 37. Krystal H: Integration and Self-Healing: Affect-Trauma-Alexi- 162:1305–1310 thymia. Hillsdale, NJ, Analytic Press, 1988 44. Herman BH, Panksepp J: Ascending endorphin inhibition of 38. Porges SW, Matthews KA, Pauls DL: The biobehavioral interface distress covalization. Science 1982; 211:1060–1062 in behavioral pediatrics. Pediatrics 1992; 90(5, part 2):789–797 45.Saxe G, Stoddard F, Courtney D, Cunningham K, Chawla N, 39. Porges SW: Emotion: an evolutionary by-product of the neural Sheridan R, King D, King L: Relationship between acute mor- regulation of the autonomic nervous system. Ann NY Acad Sci phine and the course of PTSD in children with burns. J Am 1997; 807:62–77 Acad Child Adolesc Psychiatry 2001; 40:915–921

1304 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Pathways to PTSD, Part II: Sexually Abused Children

Julie B. Kaplow, Ph.D. Objective: The goal of this research was symptoms: avoidant coping, anxiety/ to develop and test a prospective model arousal, and dissociation, all measured Kenneth A. Dodge, Ph.D. of posttraumatic stress symptoms in sexu- during or immediately after disclosure of ally abused children that includes pre- sexual abuse. Additionally, age and gen- Lisa Amaya-Jackson, M.D., M.P.H. trauma, trauma, and disclosure-related der predicted avoidant coping, while life pathways. stress and age at abuse onset predicted symptoms of anxiety/arousal. Taken to- Glenn N. Saxe, M.D., F.R.C.P. Method: At time 1, several measures were used to assess pretrauma variables, gether, these pathways accounted for ap- trauma variables, and stress reactions proximately 57% of the variance in PTSD upon disclosure for 156 sexually abused symptoms. children ages 8 to 13 years. At the time 2 Conclusions: Symptoms measured at follow-up (7 to 36 months following the the time of disclosure constitute direct, initial interview), the children were as- independent pathways by which sexually sessed for posttraumatic stress disorder abused children are likely to develop later (PTSD) symptoms. PTSD symptoms. These findings speak to Results: A path analysis involving a se- the importance of assessing children dur- ries of hierarchically nested ordinary ing the disclosure of abuse in order to least squares multiple regression analy- identify those at greatest risk for later ses indicated three direct paths to PTSD PTSD symptoms.

(Am J Psychiatry 2005; 162:1305–1310)

Child sexual abuse is an all too common event in the and resilience. The present study was designed to assess lives of children and can produce severe psychological these responses in the service of identifying a model of risk damage to victims both at the time of the abuse and years factors for PTSD symptoms in children who disclose sex- later (1, 2). Many researchers have identified posttrau- ual abuse. matic stress disorder (PTSD) as a core manifestation of Research has identified a variety of risk factors for PTSD sexual abuse trauma because of the high frequency with that are exhibited in the immediate aftermath of a trauma. which this disorder and related symptoms appear in sexu- Dissociation has been found to be particularly important ally abused children (3). A number of retrospective studies in predicting PTSD (7). Specifically, studies of adults (8–10) of adults have examined the long-term effects of child sex- and a more recent study of children, reported in this issue ual abuse on later PTSD symptoms (4, 5), but significantly of the Journal (11), have indicated that individuals who fewer have studied children to examine more immediate dissociate either during or soon after a trauma are at PTSD outcomes resulting from child sexual abuse. Even greater risk of developing PTSD. fewer have examined possible mediating mechanisms, Evidence has also suggested that anxiety/arousal re- such as the children’s reactions upon disclosure, in the de- sponses other than dissociation may predict later PTSD. velopment of later PTSD symptoms. To our knowledge, For example, in a study of adult victims of violent crime, the current study is the first to examine pretraumatic vul- Brewin et al. (12) found that a simple count of reexperi- nerabilities, trauma characteristics, and stress reactions at encing symptoms after the trauma independently pre- the time of disclosure as pathways to PTSD symptoms in dicted later PTSD. This finding is consistent with previous sexually abused children. Prospective longitudinal de- research that has linked intrusive symptoms with later signs that assess children shortly after a stressful event, problem outcomes, including PTSD (13, 14). Although such as this, are ideal for the advancement of the under- both dissociative and anxiety/arousal symptoms are in- standing of risk factors for PTSD symptoms (6). cluded in the diagnosis of acute stress disorder, consider- It is important to note that all children in the current able controversy exists over whether dissociation or anxi- study were observed while undergoing a forensic inter- ety/arousal is more predictive of PTSD (15). view, during which the child was asked to recollect and Avoidant coping is a critically important feature of the discuss potentially traumatic memories of the sexual anxiety/arousal response and has been identified as a risk abuse. As disclosure of sexual abuse can be an extremely factor for PTSD (16, 17). Retrospective studies of adult stressful event, responses immediately after such disclo- child abuse victims indicate that avoidance strategies sure may offer important information about future risk such as denial or minimizing are associated with poor psy-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1305 PTSD IN SEXUALLY ABUSED CHILDREN

TABLE 1. Pretrauma, Trauma, Disclosure, and Posttrauma in report form by the interviewers. The inclusion criteria for the Variables for 156 Sexually Abused Children current study consisted of the following: Variable Mean SD Range 1. The abuse was rated as suspicious, probable, or confirmed. Pretrauma 2. The child was between the ages of 8 and 13 years at the time Gender (male=1, female=0) 0.17 0.38 0.00–1.00 of the interview. Age at initial assessment (years) 10.70 1.80 7.92–13.92 Score for previous life stress, based 3. The interview took place between January 1998 and August on parent report of number and 2000. severity of stressful events (0=did 4. The child was from one of three counties in North Carolina. not happen, 2=major change) 6.38 4.55 0.00–16.00 Trauma: age at onset of child sexual Of the possible participants, 156 children met the criteria for abuse (years) 8.01 2.67 2.00–13.00 the study. The study group comprised 129 girls and 27 boys, with Disclosure (initial assessment) a mean age of 10.7 years (SD=1.8). Of the subjects, 56% were Afri- Avoidance (score on avoidance scale can American, 23% were Caucasian, 12% were Native American, from principal-component factor 5% were biracial, and 4% were Hispanic. Abuse was rated as con- analysis) 0.65 0.63 0.00–3.00 firmed for 54% of the participants, probable for 18%, and suspi- Dissociation (score on dissociation cious for 28%. scale of Trauma Symptom Checklist The information gathered at time 1 came from the clinicians’ for Children [22]) 8.16 5.04 0.00–20.00 Anxiety (score on anxiety scale of written reports, the videotapes, and the Trauma Symptom Check- Trauma Symptom Checklist for list for Children (22) collected at the time of the initial forensic in- Children [22]) 7.90 5.26 0.00–24.00 terview. Contact with the families for the follow-up at time 2 was Posttrauma initiated through a letter explaining the purpose of the study. Af- Interval between initial and ter complete description of the study to the subjects, written in- follow-up assessments (years) 1.57 0.69 0.58–3.00 formed consents were obtained from the parents and informed PTSD symptoms at follow-up (score assents were obtained from the children. on PTSD scale from Child Behavior Checklist [26]) 8.48 5.43 0.00–24.00 Measures The following measures, with the exception of the parent’s re- chological and psychosocial adjustment in adulthood (18, port of prior life stress and the parent’s report of the child’s PTSD symptoms, were assessed at time 1. 19). In addition, research on traumatized youth has shown that cognitive suppression and avoidance often lead to Written reports. In addition to demographic variables (i.e., age, gender), the other variable of interest in the written report follow- problems in multiple domains of functioning (20). Mea- ing the initial forensic interview was the age at onset of abuse. surement of avoidant coping strategies is complicated by These items were extracted from the written reports on the chil- the fact that, by definition, it is difficult to ascertain these dren and subsequently coded by the principal investigator symptoms from child self-reports (21). In the current (J.B.K.) and a trained research assistant. The reliability estimates were satisfactory, with kappas ranging from 0.59 to 0.90. Discrep- study we surmounted this obstacle by using a behavioral ant ratings were discussed by the principal investigator and the coding system to rate the children’s avoidant behaviors research assistant until a consensus was reached. during videotaped forensic interviews. Videotaped interviews. Semistructured investigative inter- The current study utilized path analytic techniques to views were conducted by skilled clinicians trained in the evalua- evaluate the ways in which pretrauma variables, trauma tion of sexually abused children, and all interviews were videotaped. A detailed coding system was developed and used to characteristics, and stress reactions upon disclosure of analyze the child’s emotional and behavioral responses through- sexual abuse are related to later PTSD symptoms in chil- out the interview, which would be operationalized as child coping dren. As described, anxiety and dissociative reactions in variables in the study. The interviews of the participants selected the wake of a trauma have been identified as important for interrater reliability were randomly sampled intermittently throughout the coding period, and agreement between each of components of the acute trauma response, although there the two coders’ ratings and those of the principal investigator was is considerable controversy over which is more strongly estimated for 40% of the study participants. The reliability esti- related to deleterious long-term outcomes. Consequently, mates were satisfactory, with kappas ranging from 0.63 to 0.82 a main goal of the current study was to assess the relative and Pearson’s correlation coefficients ranging from 0.41 to 0.98. Because of the high correlations among many of the items, a importance of dissociation and anxiety in the immediate principal-component factor analysis was conducted. The most aftermath of the disclosure of sexual abuse. readily interpretable loading pattern was found by using varimax rotation. Each item was assigned to the factor on which it had the highest loading, and no item was allowed to contribute to more Method than one factor. The scores for the items on each scale were summed and averaged, with higher scores reflecting higher levels Participants and Procedure of the particular construct. The analysis identified 14 behaviors The participants in the current study had been referred to a that factored onto three scales. Of interest to the current study is treatment facility that offers services to children considered pos- the avoidance scale, consisting of four items: 1) attempting to dis- sible victims of sexual abuse. Each of the participants had been tract the interviewer, 2) appearing avoidant (e.g., by not answer- medically examined, interviewed, and videotaped for forensic ing questions), 3) appearing distracted, and 4) appearing fidgety purposes. The occurrence of sexual abuse was rated by a multi- (alpha=0.79). disciplinary evaluation team as confirmed, probable, suspicious, Anxiety and dissociation. The Trauma Symptom Checklist for unknown, or “no evidence.” All of the interviews were written up Children (22) is a 54-item self-report instrument that evaluates

1306 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KAPLOW, DODGE, AMAYA-JACKSON, ET AL.

TABLE 2. Correlations Among Pretrauma, Trauma, Disclosure, and Posttrauma Variables for 156 Sexually Abused Children Correlation (r) Age at Onset Interval Between PTSD Variable Life Stress of Abuse Avoidance Dissociation Anxiety Assessments Symptoms Pretrauma Age at initial assessment 0.06 0.44** –0.22* 0.11 –0.01 –0.39** 0.03 Previous life stress 0.29* 0.15 –0.06 0.14 –0.15 0.20 Trauma: age at onset of child sexual abuse –0.17 –0.06 –0.14 –0.20 –0.08 Disclosure (initial assessment) Avoidance –0.11 –0.04 0.01 0.24* Dissociation 0.68** 0.10 0.67** Anxiety –0.16 0.52** Posttrauma Interval between initial and follow-up assessments –0.04 PTSD symptoms at follow-up *p<0.05. **p<0.01. posttraumatic symptoms, including anxiety/arousal symptoms ance, and dissociation) that together accounted for 57% of the and dissociation, in children and adolescents ages 8 to 16. Various variance in PTSD symptoms (R2=0.57). Once these variables were studies that have used the checklist indicate that it is reliable chosen, we began to include antecedent variables that would al- (with alpha values in the mid to high 80s for all scales) and has low anxiety, avoidance, and dissociation to serve as potential me- convergent and predictive validity in groups of both traumatized diators. In order to constrain the number of paths in this model, and nontraumatized children and adolescents (23, 24). Dissocia- bivariate relationships between any variable and PTSD that did tion in this study was measured by the 10-item dissociation scale not reach conventional levels of significance (p<0.05) were de- of the Trauma Symptom Checklist for Children, and anxiety was leted. In addition, the model was further constrained by temporal measured by the nine-item anxiety scale. relationships. The final path analytic model provided excellent fit Previous life stress. At the follow-up interview, the parent was indices (comparative fit index=1.00, Tucker-Lewis index=1.00, asked to report on the number of stressful events that had taken root mean square error of approximation=0.00). place during the child’s life and how each event may have affected the child. Each item was rated on a scale from 0 to 2 (0=did not Results happen, 1=minor change, 2=major change). A total life stress score was derived from the sum of all of the items. Preliminary Analyses Parent report of child posttraumatic stress. At the follow- up interview, the parent also completed the Child Behavior Table 1 displays descriptive statistics for all variables of Checklist (25), a parent-report measure of the child’s internalizing interest to the current study. A series of t tests and analyses and externalizing symptoms. The PTSD scale (26), derived from of variance revealed no significant differences in time 2 19 items on the Child Behavior Checklist, has been shown to have PTSD symptoms as a function of gender, race, or county. good internal consistency (alpha=0.86; Becker, unpublished 2000 paper) and is highly correlated with the total PTSD scale of the Table 2 presents Pearson’s correlation coefficients for Trauma Symptom Checklist for Young Children (27) (r=0.69). the associations among pretrauma and trauma variables, disclosure reactions, and later PTSD symptoms. Data Analysis In order to make use of all data available, we approached the Path Analysis issue of missing data at time 2 by using maximum likelihood estimation techniques for missing data (28). Evidence suggests that Figure 1 illustrates the results of the path analysis, with there is far less risk of bias with retention of missing cases and partial correlation coefficients (beta weights) given for use of this method than with listwise deletion in complete case each path remaining after nonsignificant paths were reanalysis (29). (For a thorough review of methods for handling moved. The results indicate three direct paths to PTSD missing data, see the discussion by Graham et al. [30].) The cur- symptoms from the disclosure reactions: avoidance, disso- rent study used a path analytic strategy similar to that used by Shalev et al. (31) in their prospective study of acutely traumatized ciation, and anxiety. They also show four indirect paths to individuals. As this strategy follows a prospective longitudinal PTSD symptoms from pretrauma and trauma variables: method, the directionality of paths is constrained by the time at age, gender, life stress, and age at onset of abuse. More spe- which the variables are assessed. Accordingly, we divided vari- cifically, anxiety/arousal served as a mediator between life ables into four categories: 1) PTSD symptoms (dependent vari- stress and PTSD and between age at onset and PTSD. In ad- able), 2) disclosure reactions (variables measured upon disclosure of abuse), 3) trauma variables (variables specific to the dition, avoidance served as a mediator between age and trauma itself), and 4) pretrauma variables (variables describing PTSD and between gender and PTSD. Dissociation was not the child prior to the trauma). only a direct predictor of PTSD but also an indirect predic- A series of hierarchically nested ordinary least squares multiple tor of PTSD symptoms by way of anxiety symptoms. Taken regression analyses were used to estimate direct and indirect ef- together, these pathways accounted for 57% of the vari- fects among variables. Guided by theory and previous empirical ance in PTSD. Given the high explanatory value of these research, we chose combinations of variables that accounted for a high percentage of the variance in PTSD symptoms (high R2). pathways, the model helps to shed light on the relatively Accordingly, we chose three primary variables (i.e., anxiety, avoid- unknown course of PTSD in sexually abused children.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1307 PTSD IN SEXUALLY ABUSED CHILDREN

FIGURE 1. Path Analytic Model for the Development of dissociation and the parasympathetically mediated a PTSD in 156 Sexually Abused Children “freeze” or “immobilization” response (37–40). Although the current study did not directly measure these biobehav- Anxiety/ ioral systems, i.e., sympathetic and parasympathetic activ- Life 0.33 arousal stress ity, the independent contributions made by the dissocia- –0.22 0.22 tion and anxiety/arousal pathways in predicting PTSD Age at onset 0.71 suggest independence of the biobehavioral processes. Further, as dissociation by this theory is presumed to be 0.22 0.26 Gender Avoidance PTSD a more primitive response that occurs only after the fight- or-flight arousal system has been overwhelmed, it makes –0.21 sense that there is a unidirectionality of the relationship 0.54 between dissociation and anxiety/arousal found in this Age Dissociation study. Dissociation is significantly associated with anxiety/arousal, whereas anxiety/arousal is not significantly associated with dissociation. In other words, if a child must initially experience arousal/anxiety to reach a disso- Pretrauma Trauma Disclosure Posttrauma ciated state, then children with dissociation would neces- a The values in the model are partial correlation coefficients (beta sarily experience some symptoms of arousal/anxiety. weights). R2=0.57. Anxiety/arousal symptoms following trauma have been identified in the adult literature as critical in predicting Discussion later PTSD (12, 41); however, almost no studies to date have examined the predictive utility of these symptoms in To our knowledge, this is the first study to examine the relation to later PTSD in children. The results of the current utility of a prospective model incorporating pretrauma, study demonstrate that anxiety/arousal symptoms imme- trauma, and disclosure variables in pathways leading to diately after disclosure of sexual abuse make a unique con- PTSD in sexually abused children. The results of this tribution to the development of PTSD symptoms. study demonstrate that sexually abused children who ex- It is noteworthy that life stress positively predicted anx- hibit symptoms of avoidance, anxiety/arousal, or dissoci- iety/arousal symptoms. In addition, age at onset of the ation either during or immediately following disclosure of abuse negatively predicted anxiety/arousal symptoms, in- abuse are at increased risk of developing PTSD symptoms dicating that the earlier the abuse began, the more likely at a later date. These findings speak to the importance of the child was to demonstrate these symptoms at the time assessing children’s reactions as soon as possible after of the disclosure. Rutter (42) hypothesized that the accu- disclosure. mulation of multiple stressors in children’s lives dramati- Given the overwhelming stress experienced by sexual cally increases the risk of permanent developmental dam- abuse victims during the time of the event as well as dur- age and the manifestation of PTSD symptoms. ing disclosure, it is not surprising that childhood sexual Although children’s use of coping strategies has been abuse is strongly associated with dissociative responses (7, identified as a particularly promising area of study (43), 32, 33). However, because of previous reliance on retro- this field of inquiry has been hampered by the traumatic spective reports of adult sexual abuse victims, the pro- nature of sexual abuse and its legal and clinical implica- spective relation between dissociative symptoms and later tions. We believe that this study is the first to demonstrate PTSD symptoms in childhood has remained unclear. This a positive relation between avoidant coping immediately study serves as confirmation that children who report dis- after the discovery of trauma and later PTSD symptoms in sociative symptoms immediately after disclosure of abuse sexually abused children. These findings are consistent are at greater risk for later PTSD symptoms. In fact, disso- with studies of adults suggesting that strategies such as ciation appeared to be the strongest predictor of PTSD denial, minimization, or purposeful forgetting are associ- symptoms in this group of children. It has been theorized ated with greater psychological difficulties in the long that dissociative responses may prevent the open expres- term (19, 44). sion of emotions and cognitions associated with the The model also demonstrates that boys are more likely trauma, which is likely to lead to insufficient processing of to exhibit avoidant behaviors upon disclosure. This find- the trauma, more reexperiencing symptoms, and conse- ing is consistent with a number of studies showing higher quently, worse PTSD symptoms (34–36). levels of emotional expression in girls (45, 46) and women Dissociation may also represent the phenotype of a (47) than in boys and men, respectively. The current study biobehavioral vulnerability to traumatic events. A number also indicated that younger children are more likely to be of researchers have drawn connections between the anxi- avoidant upon disclosure. Peterson (48) described coping ety/hyperarousal symptoms and the sympathetically me- as a process that is significantly influenced by develop- diated fight-or-flight response and between symptoms of ment. Younger children are less likely to have the language

1308 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 KAPLOW, DODGE, AMAYA-JACKSON, ET AL. capacity and emotion identification abilities that older References children have, thereby making it more difficult for them to 1. Carlson EB, Furby L, Armstrong J, Shlaes J: A conceptual frame- express their thoughts and feelings. work for the long-term psychological effects of traumatic childhood abuse. Child Maltreat 1997; 2:272–295 Limitations 2. Fergusson DM, Mullen P (eds): Childhood Sexual Abuse: An Ev- Although the current study exhibits important strengths, idence Based Perspective. Thousand Oaks, Calif, Sage Publica- tions, 1999 there are several limitations that should be noted. The 3. Kendall-Tackett KA, Williams LM, Finkelhor D: Impact of sexual findings generalize to the population of children identi- abuse on children: a review and synthesis of recent empirical fied by county social service departments as possibly vic- studies. Psychol Bull 1993; 113:164–180 timized by sexual abuse, but their generalizability to the 4. Roesler TA, McKenzie N: Effects of childhood trauma on psy- total population of abused children is unclear. It is also chological functioning in adults sexually abused as children. J Nerv Ment Dis 1994; 182:145–150 possible that the forensic interview represented the first 5. Widom CS: Posttraumatic stress disorder in abused and ne- disclosure of abuse for some children, whereas other chil- glected children grown up. Am J Psychiatry 1999; 156:1223– dren may have disclosed it in great detail to other people 1229 prior to the interview. The findings of studies on the rela- 6. Yehuda R: Risk Factors for Posttraumatic Stress Disorder. Wash- tionship between disclosure and psychological function- ington, DC, American Psychiatric Press, 1999 ing have been equivocal (49). Consequently, one must 7. Putnam FW: Dissociation in Children and Adolescents: A Devel- opmental Perspective. New York, Guilford Press, 1997 consider the possibility that the child’s reactions observed 8. Marmar CR, Weiss DS, Schlenger WE, Fairbank JA, Jordan BK, upon disclosure are mainly a response to the interview it- Kulka RA, Hough RL: Peritraumatic dissociation and posttrau- self and are unrelated to the child’s attempt to cope with matic stress in male Vietnam theater veterans. Am J Psychiatry traumatic reminders. 1994; 151:902–907 9. Dancu CV, Riggs DS, Hearst-Ikeda D, Shoyer BG, Foa EB: Disso- Clinical Implications ciative experiences and posttraumatic stress disorder among female victims of criminal assault and rape. J Trauma Stress This research has crucial implications for all providers 1996; 9:253–267 conducting clinical or forensic evaluations and treatments 10. Koopman C, Classen C, Spiegel D: Predictors of posttraumatic for sexually abused children by helping to identify chil- stress symptoms among survivors of the Oakland/Berkeley, dren who appear to be the most at risk for developing Calif, firestorm. Am J Psychiatry 1994; 151:888–894 11. Saxe G, Stoddard F, Hall E, Chawla N, Lopez C, Sheridan R, King PTSD symptoms. The findings lend support to the notion D, King L, Yehuda R: Pathways to PTSD, part I: children with that children who exhibit avoidance, anxiety/arousal, and burns. Am J Psychiatry 2005; 161:1299–1304 especially, dissociation during disclosure are likely to need 12. Brewin CR, Andrews B, Rose S, Kirk M: Acute stress disorder treatment in order to prevent the development of later and posttraumatic stress disorder in victims of violent crime. PTSD symptoms. Coping is by definition a consciously Am J Psychiatry 1999; 156:360–366 13. Joseph S, Yule W, Williams R: Emotional processing in survivors chosen means of dealing with stress and thereby amena- of the Jupiter cruise ship disaster. Behav Res Ther 1995; 33: ble to modification. Therefore, coping strategies such as 187–192 avoidance are directly relevant as intervention targets in 14. McFarlane AC: Avoidance and intrusion in post-traumatic stress therapy (50). Consistent with reports of the outcomes of disorder. J Nerv Ment Dis 1992; 180:439–445 PTSD treatment (34, 51), the current findings suggest that 15. Marshall RD, Spitzer R, Liebowitz MR: Review and critique of the new DSM-IV diagnosis of acute stress disorder. Am J Psychi- treatments encouraging open expression of thoughts and atry 1999; 156:1677–1685 feelings surrounding the abuse can help to deter growth in 16. Shapiro DL, Levendosky AA: Adolescent survivors of childhood PTSD symptoms over time. Further, the relative indepen- sexual abuse: the mediating role of attachment style and cop- dence of dissociative and anxiety/arousal pathways to ing in psychological and interpersonal functioning. Child PTSD symptoms suggests the possibility for targeted treat- Abuse Negl 1999; 23:1175–1191 ments aimed at particular biobehavioral systems. 17. Tremblay C, Hebert M, Piche C: Coping strategies and social support as mediators of consequences in child sexual abuse victims. Child Abuse Negl 1999; 23:929–945 Received Dec. 12, 2003; revision received July 13, 2004; accepted 18. Erickson MF, Egeland B: A developmental view of the psycho- Aug. 2, 2004. From the Department of Psychology and the Center for logical consequences of maltreatment. School Psychol Rev Child and Family Health, Duke University, Durham, N.C.; and the De- 1987; 16:156–168 partment of Child and Adolescent Psychiatry and the Center for Med- 19. Varia R, Abidin RR, Dass P: Perceptions of abuse: effects on ical and Refugee Trauma, Boston University Medical Center. Address correspondence and reprint requests to Dr. Kaplow, Department of adult psychological and social adjustment. Child Abuse Negl Psychiatry, New Jersey Medical School, University of Medicine and 1996; 20:511–526 Dentistry of New Jersey, P.O. Box 1709, Newark, NJ 07101-1709; 20. March JS, Amaya-Jackson L, Terry R, Costanzo P: Post-traumatic [email protected] (e-mail). symptomatology in children and adolescents after an indus- The authors thank Melanie J. Bonner, E. Jane Costello, John F. Curry, trial fire in Hamlet, North Carolina. J Am Acad Child Adolesc and Susan Roth for their comments and Amber Diagre, Brian Maxey, Psychiatry 1997; 36:1080–1089 Taeh Ward, Jeanne Arnts, Donna Potter, Jennifer Lansford, Dan King, 21. Practice parameters for the assessment and treatment of chil- and Lynda King for their contributions to this study. dren and adolescents with posttraumatic stress disorder. J Am Acad Child Adolesc Psychiatry 1998; 37(Oct suppl):4S–26S

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1309 PTSD IN SEXUALLY ABUSED CHILDREN

22. Briere J: Trauma Symptom Checklist for Children (TSCC) Profes- 37. Perry BD: Incubated in terror: neurodevelopmental factors in sional Manual. Odessa, Fla, Psychological Assessment Re- the “cycle of violence,” in Children in a Violent Society. Edited sources, 1996 by Osofsky J. New York, Guilford, 1997, pp 124–148 23. Crouch JL, Smith DW, Ezzell CE, Saunders BE: Measuring reac- 38. Perry BD, Pollard RA, Blakley TL, Baker WL, Vigilante D: Child- tions to sexual trauma among children: comparing the Chil- hood trauma, the neurobiology of adaptation, and use-depen- dren’s Impact of Events Scale and the Trauma Symptom Check- dent development of the brain: how states become traits. In- list for Children. Child Maltreat 1999; 4:255–263 fant Ment Health J 1995; 16:271–291 24. Friedrich WN, Jaworski TM, Huxsahl JE, Bengtson BS: Dissocia- 39. Porges S: An evolutionary by-product of the neural regulation tive and sexual behaviors in children and adolescents with sex- of the autonomic nervous system. Ann NY Acad Sci 1997; 807: ual abuse and psychiatric histories. J Interpers Violence 1997; 62–77 12:155–171 40. Porges SW: Orienting in a defensive world: mammalian modi- 25. Achenbach TM: Manual for the Youth Self-Report and 1991 fications of our evolutionary heritage: a polyvagal theory. Psy- Profile. Burlington, University of Vermont, Department of Psy- chophysiology 1995; 32:301–318 chiatry, 1991 41. Harvey AG, Bryant RA: Two-year prospective evaluation of the 26. Wolfe V, Gentile C, Wolfe DA: The impact of sexual abuse on relationship between acute stress disorder and posttraumatic children: a PTSD formulation. Behav Ther 1989; 20:215–228 stress disorder following mild traumatic brain injury. Am J Psy- 27. Briere J, Johnson K, Bissada A, Damon L, Crouch JL, Gil E, Han- chiatry 2000; 157:626–628 son R, Ernst V: The Trauma Symptom Checklist for Young Chil- 42. Rutter M: Continuities and discontinuities from infancy, in dren (TSCYC): reliability and association with abuse exposure in Handbook of Infant Development, 2nd ed. Edited by Osofsky J. a multi-site study. Child Abuse Negl 2001; 25:1001–1014 New York, John Wiley & Sons, 1987, pp 1256–1296 43. Valle LA, Silovsky JF: Attributions and adjustment following 28. Muthen LK, Muthen BO: M Plus Statistical Analysis With Latent child sexual and physical abuse. Child Maltreat 2002; 7:9–25 Variables: User’s Guide. Los Angeles, Muthén & Muthén, 1998 44. Himelein MJ, McElrath JAV: Resilient child sexual abuse survi- 29. Curran PJ, Hussong AM: The use of latent trajectory models in vors: cognitive coping and illusion. Child Abuse Negl 1996; 20: psychopathology research. J Abnorm Psychol 2003; 112:526– 747–758 544 45. Brody LR: Gender, emotional expression, and parent-child 30. Graham J, Cumsille P, Elek-Fiske E: Methods for Handling Miss- boundaries, in Emotion: Interdisciplinary Perspectives. Edited ing Data. State College, Pennsylvania State University, Method- by Kavanaugh RD, Zimmerberg B. Hillsdale, NJ, Lawrence Erl- ology Center, 2001, pp 1–76 baum Associates, 1996, pp 139–170 31. Shalev AY, Peri T, Canetti L, Schreiber S: Predictors of PTSD in in- 46. Polce-Lynch M, Myers BJ, Kliewer W, Kilmartin C: Adolescent jured trauma survivors: a prospective study. Am J Psychiatry self-esteem and gender: exploring relations to sexual harass- 1996; 153:219–225 ment, body image, media influence, and emotional expres- 32. Kirby JS, Chu JA, Dill DL: Correlates of dissociative symptoma- sion. J Youth Adolesc 2001; 30:225–244 tology in patients with physical and sexual abuse histories. 47. Sells DJ, Martin RB: Gender and modality differences in experi- Compr Psychiatry 1993; 34:258–263 encing and emotional expression. Canadian J Counseling 2001; 33. Zlotnick C, Shea MT, Pearlstein T, Begin A, Simpson E, Costello 35:176–188 E: Differences in dissociative experiences between survivors of 48. Peterson L: Coping by children undergoing stressful medical childhood incest and survivors of assault in adulthood. J Nerv procedures: some conceptual, methodological, and therapeu- Ment Dis 1996; 184:52–54 tic issues. J Consult Clin Psychol 1989; 57:380–387 34. Amir N, Stafford J, Freshman M, Foa E: Relationship between 49. Nagel DE, Putnam FW, Noll JG: Disclosure patterns of sexual trauma narratives and trauma pathology. J Trauma Stress abuse and psychological functioning at a 1-year follow-up. 1998; 11:385–392 Child Abuse Negl 1997; 21:137–147 35. Deblinger E, McLeer SV, Henry D: Cognitive behavioral treat- 50. Chaffin M, Wherry JN, Dykman R: School age children’s coping ment for sexually abused children suffering post-traumatic with sexual abuse: abuse stresses and symptoms associated stress: preliminary findings. J Am Acad Child Adolesc Psychiatry with four coping strategies. Child Abuse Negl 1997; 21:227– 1990; 29:747–752 240 36. Pennebaker JW: Putting stress into words: health, linguistic, 51. Jaycox LH, Morral AR, Foa EB: The influence of emotional en- and therapeutic implications. Behav Res Ther 1993; 31:539– gagement and habituation on outcome of exposure therapy 548 for PTSD. J Consult Clin Psychol 1998; 66:185–192

1310 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Improving the Rates of Quitting Smoking for Veterans With Posttraumatic Stress Disorder

Miles McFall, Ph.D. Objective: Smoking is highly prevalent Results: Subjects assigned to integrated and refractory among people with post- care were five times more likely than sub- Andrew J. Saxon, M.D. traumatic stress disorder (PTSD). This study jects undergoing the usual standard of care aimed to improve the rate of quitting to abstain from smoking across follow-up assessment intervals (odds ratio=5.23). smoking for veterans with PTSD by inte- Charles E. Thompson, M.D. Subjects in the integrated care condition grating treatment for nicotine dependence were significantly more likely than subjects into mental health care. Dan Yoshimoto, M.A. in usual standard of care to receive trans- Method: Smokers undergoing treatment dermal nicotine and nicotine gum. They Carol Malte, M.S.W. for PTSD (N=66) were randomly assigned also received a greater number of smok- to 1) tobacco use treatment delivered by ing-cessation counseling sessions. Stopping smoking was not associated with worsen- mental health providers and integrated Kristy Straits-Troster, Ph.D. ing symptoms of PTSD or depression. with psychiatric care (integrated care) ver- Conclusions: Smoking-cessation inter- Evan Kanter, M.D., Ph.D. sus 2) cessation treatment delivered separately from PTSD care by smoking-cessa- ventions can be safely incorporated into routine mental health care for PTSD and tion specialists (usual standard of care). Xiao-Hua Andrew Zhou, Ph.D. are more effective than treatment deliv- Seven-day point prevalence abstinence ered separately by a specialized smoking- was the primary outcome, measured at 2, Cynthia M. Dougherty, A.R.N.P., cessation clinic. Integrating cessation 4, 6, and 9 months after random assign- treatment into psychiatric care may have Ph.D. ment. Data were analyzed by using a gen- the potential for improving smoking quit eralized estimating equations approach rates in other populations of chronically Bonnie Steele, R.N., Ph.D. following the intent-to-treat principle. mentally ill smokers.

(Am J Psychiatry 2005; 162:1311–1319)

Posttraumatic stress disorder (PTSD) is one of the most clinical trials (12, 14). However, these treatments are only prevalent mental disorders (1, 2), particularly among Vet- as useful as the ability of health care organizations to de- erans Administration (VA) health care enrollees (3, 4). In liver them effectively to individuals who need them most. the general population, PTSD is associated with high rates Primary care providers only infrequently apply even brief, of smoking (45% [5] versus the national average of 23% [6]) cost-effective tobacco-cessation interventions to smokers and a fourfold increased risk for nicotine dependence (7). (15–18), despite the fact that 60%–70% of smokers want to The rates of smoking in veterans with PTSD (53%–66% [8– quit (6, 11). Nicotine-dependence treatment in patients 10]) are approximately double those of VA enrollees in with mental disorders may be particularly neglected since general (30% [11]). Moreover, a greater proportion of psychiatric patients receive cessation counseling at only smokers with combat-related PTSD smoke heavily (>25 38% of primary care visits and 12% of visits with a psychi- cigarettes per day) compared with veterans who smoke atrist (16). Referral of patients to specialty smoking-cessa- but do not have PTSD (48% versus 28% [8]). PTSD is asso- tion clinics is a commonly used alternative to primary ciated with a smoking quit rate of only 23% (5), about half care-based delivery of tobacco use treatment. However, that of lifetime smokers without a mental disorder (12), the effectiveness of these clinics is compromised by poor and falls third from the bottom in a ranking of quit rates patient compliance, with rates of attendance as low as for 13 mental disorders (5). Smokers with PTSD also expe- 13%–14% (19, 20) and limited capacity to provide repeated rience nicotine withdrawal symptoms in response to intervention for a chronic, relapsing disorder such as nic- encounters with trauma-related stimuli (13) and report otine dependence. These limitations create a situation in smoking in order to relieve anxiety and tension (8). Taken which only 17% of smokers in the nation’s largest health together, this research suggests a dynamic relationship be- care system (the VA) report receiving desired cessation tween PTSD and tobacco use that argues for a coordinated treatment in the previous year (11). approach to the treatment of both disorders. The simultaneous treatment of two or more interwoven Several pharmacological and behavioral treatments for disorders by a single provider team has shown promising nicotine dependence have shown efficacy in controlled clinical effectiveness for patients with severe mental ill-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1311 PTSD AND QUITTING SMOKING

TABLE 1. Baseline Characteristics of Subjects Receiving Integrated Care Versus the Usual Standard of Carea Veterans Receiving Integrated Care Veterans Receiving Usual Standard of Care Variable N % Mean SD N % Mean SD Age (years) 33 52.9 11.1 33 52.3 8.1 Male sex 29 88 32 97 White race 26 79 31 94 Married 18 55 14 44 Unemployed 27 82 27 84 Receiving Department of Veterans Affairs disability 29 88 33 79 Number of current axis I disorders 33 2.0 0.7 33 1.9 0.6 PTSD Checklist score 33 63.7 13.1 33 63.1 13.0 Beck Depression Inventory score 33 29.6 9.5 33 27.4 10.7 Cigarettes smoked/day 33 25.2 11.8 33 29.3 15.8 Fagerstrom Test for Nicotine Dependence score 33 5.9 2.1 32 6.4 2.0 Expired carbon monoxide (ppm) 33 24.5 10.4 33 27.6 13.7 Years spent smoking 33 33.4 13.4 29 29.3 9.5 Living with other smokers 14 42 12 41 Confidence in quittingb 33 2.3 1.1 29 2.7 1.0 a There were no statistically significant differences between study conditions on baseline characteristics (all p>0.05). b Ratings were made on a 5-point Likert scale ranging from 0=“not at all” to 4=“extremely.” ness and comorbid substance use disorders (21, 22). This order other than tobacco use. After providing a complete descrip- study accordingly tested an approach for improving the tion of the study, written informed consent was obtained. The effectiveness of tobacco-cessation service delivery by in- subjects were then randomly assigned to integrated care (N=33) versus the usual standard of care (N=33) for smoking cessation. tegrating treatment for smoking into the routine mental The rate of refusal to participate in this study was 3% for help- health care of patients with PTSD. Specifically, this ran- seeking smokers. domized, controlled clinical trial compared the effective- Treatment Conditions ness of two different methods for delivering guideline- based smoking-cessation treatment to patients receiving All study subjects received psychotropic medications for PTSD throughout their participation in the study, as prescribed by two VA mental health care: 1) brief smoking-cessation inter- PTSD clinic psychiatrists and a nurse practitioner. Each subject ventions integrated with ongoing mental health care and also received psychotherapy from an assigned case manager who delivered by mental health providers (integrated care) ver- coordinated their mental health care. Case managers included sus 2) smoking-cessation interventions delivered sepa- four psychologists, one social worker, an addictions therapist, rately from mental health care by smoking-cessation spe- and a technician. cialists (usual standard of care). It was hypothesized that Integrated Care: Experimental Condition integrated care would result in higher rates of smoking The subjects randomly assigned to integrated care received cessation than would usual standard of care. A secondary smoking-cessation interventions administered by their assigned objective of this study was to determine if smoking cessa- PTSD clinic prescriber and case manager. Integrated care was tion was associated with worsening PTSD or depression modeled after the brief clinical interventions for primary care symptoms. Anticipating this possibility is important, practitioners published in the U.S. Public Health Service’s clinical practice guideline titled Treating Tobacco Use and Dependence given the high co-occurrence of depression with PTSD (14). PTSD clinic staff received approximately 3 hours of training, (23) and prior research showing that smoking abstinence plus as-needed consultation in smoking-cessation treatment from can exacerbate depression in individuals with a history of the clinic director (M.M.). They then delivered interventions using this disorder (24–27). a treatment manual (available upon request from the first author) that operationalized interventions for each session. The subjects received smoking-cessation protocol medications Method (bupropion, transdermal nicotine, nicotine polacrilex gum, and nicotine spray) from the psychiatrist or nurse practitioner man- Subjects aging their pharmacological treatment of PTSD. Prescribers in the Subjects (N=66) were recruited from the VA Puget Sound Health integrated care and the usual standard of care conditions agreed Care System PTSD clinic, which provides specialized outpatient to adopt the practice of routinely prescribing bupropion, trans- treatment for chronic PTSD. Table 1 reports demographic and dermal nicotine, and nicotine gum or spray in order to standard- smoking characteristics for the study group as well as psychomet- ize the medication protocol across conditions. However, the pre- ric data showing moderate to severe levels of symptoms on the scribers were allowed to use their discretion to select only some of PTSD Checklist (28) and the Beck Depression Inventory (29). Sub- these medications, depending on patient preferences and medi- jects in the integrated care and usual standard of care conditions cal contraindications. did not differ significantly on baseline characteristics (all p values The core behavioral counseling components of integrated care, >0.05). administered by case managers, consisted of 1) the assessment of Subjects were included if they smoked ≥10 cigarettes per day, tobacco use status, abstinence history, and individualized rea- expressed a willingness to receive smoking-cessation treatment, sons for quitting; 2) education about the health risks of smoking and met DSM-IV criteria for PTSD. Exclusion criteria were 1) the and the benefits of quitting; 3) advice to quit smoking; 4) applica- use of smokeless tobacco, pipes, or cigars, 2) the presence of un- tion of motivational interventions for ambivalent smokers; 5) set- stable axis I disorders, and 3) current substance dependence dis- ting a date to quit; 6) behavioral counseling to help subjects pre-

1312 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 McFALL, SAXON, THOMPSON, ET AL. pare to quit smoking, coping with smoking urges and barriers to Smoking outcomes. Seven-day point prevalence abstinence, abstinence, and developing problem-solving skills; 7) self-help verified by expired carbon monoxide (≤10 ppm), was the primary reading materials; 8) intrasession support and assistance in iden- outcome measure used to compare integrated care and the usual tifying extrasession social support; and 9) self-directed behavioral standard of care at 2, 4, 6, and 9 months after random assignment. methods for reducing anxiety, consisting of a relaxation training The period after randomization provided a consistent marker tape and written materials on stress management. The protocol across subjects of the initiation of the treatment process. The required case managers to administer five individual behavioral convention of using an end-of-treatment phase or an initial quit counseling sessions on a once-weekly basis, plus one follow-up date as a starting point for abstinence measurements was not fol- contact (the sessions averaged 20 minutes each). These interven- lowed because this study aimed to compare tobacco cessation tions were rolled into regularly scheduled visits addressing PTSD treatments as they are actually practiced in a clinical setting. Spe- and comorbid mental disorders for subjects whose treatment cifically, guideline-adherent treatment for a chronic relapsing plan ordinarily included individual sessions with a case manager. condition, such as tobacco use disorder, is a continuous process For subjects receiving only group therapy for PTSD, case manag- requiring multiple quit attempts and reapplication of interven- ers scheduled separate individual smoking-cessation sessions. tions aimed at “recycling” patients who relapse (14, 30, 31). Con- After delivering the six core behavioral counseling sessions, clini- sistent with expert recommendations (12, 14, 31), study treat- cians used their discretion to periodically assess smoking status ments were not defined by a discrete episode of care marked by a and reinstate cessation treatment for subjects who relapsed. fixed endpoint, number of sessions, or single quit date. Repeated 7-day point prevalence abstinence was used as a Usual Standard of Care: Comparison Condition proxy for prolonged abstinence, consistent with prior research The subjects randomly assigned to the usual standard of care (32–36). Repeated 7-day point prevalence abstinence was based were referred to the VA Puget Sound Health Care System’s Smok- on three consecutive follow-up assessment intervals (4, 6, and 9 ing Cessation Clinic. Nursing personnel who staffed this clinic months after random assignment). The 2-month assessment was were trained at Mayo Clinic’s Nicotine Dependence Treatment not included in this computation, allowing subjects a stabiliza- Center and had extensive experience in cessation treatment. This tion period to recover from lapses after initial intervention (37). clinic delivered U.S. Public Health Service guideline-adherent Although repeated 7-day point prevalence abstinence is a less cessation treatment (14). Clinic providers used the same algo- conservative measure than prolonged abstinence, the two mea- rithm as integrated care providers for prescribing smoking-cessa- sures nevertheless correlate highly (r=0.85–0.94 [38, 39]). tion medications. The subjects attended one group orientation Mental health quality assurance outcomes. The PTSD Check- class, followed by individual sessions in which they received med- list (28) and the Beck Depression Inventory (29) were completed ications and behavioral counseling. Unrestricted access to usual by subjects at baseline and 6- and 9-month follow-ups. The sub- standard of care treatment sessions was provided to all subjects. jects rated symptoms that occurred during the prior week. The That is, the number of treatment contacts received was deter- PTSD Checklist and the Beck Depression Inventory were used to mined by clinical recommendations of usual standard of care assess changes in mental health symptoms over time in order to providers and the preferences and initiative of subjects. The sub- determine if smoking cessation was associated with worsening jects assigned to the usual standard of care condition received ab- symptoms. solutely no tobacco-cessation interventions from their PTSD clinic providers. Patient satisfaction outcomes. The subjects in each study condition rated their satisfaction with the amount and quality of Measures smoking-cessation treatment received by using a 4-point Likert Delivery of study treatments. The Consumer Health Informa- scale (1=“very dissatisfied” and 4=“very satisfied”). The ratings tion and Performances Sets, an administrative database account- were gathered at the terminal (9-month) study assessment. ing for services delivered by the VA, was used to track smoking- Data Analysis cessation medications actually received by the subjects. This database provided information about the proportion of subjects Abstinence data were analyzed in an intent-to-treat analysis by who filled prescriptions for study medications as well as doses for using marginal logistic regression with the method of generalized these medications. The number of smoking cessation behavioral estimating equations. This statistical approach to analyzing smok- counseling sessions delivered to the subjects was extracted from ing-cessation outcome data conforms to recommendations by an the VA’s computerized patient record system, where clinicians are expert panel from the Society for Research on Nicotine and required to document all patient visits. Tobacco (39). The generalized estimating equation method has Adherence of integrated care case managers to the treatment several advantages in that it accounts for the correlation of re- manual was assessed by independent review of the computerized peated measurement, uses all available data points, and is sensi- patient record system. These records contained provider docu- tive to the pattern of change over time (40). Covariates in the anal- mentation of interventions delivered during each visit on spe- ysis were baseline measures of severity of nicotine dependence cially prepared charting templates that listed a total of 31 specific (the Fagerstrom Test for Nicotine Dependence [41]) and depres- protocol interventions prescribed by the manual. These interven- sion (the Beck Depression Inventory). Depression was included as tions constituted the core behavioral counseling components of a covariate because of evidence showing that it may adversely af- integrated care (e.g., “set and recorded a quit date,” “identified fect smoking-cessation outcomes (42–44). strategies for coping with withdrawal symptoms”). After each ses- The effect of smoking cessation on mental health symptoms at sion, case managers checked off interventions delivered on the 6- and 9-month follow-up assessments was analyzed by comput- charting template and provided a narrative summary of smoking ing change from baseline scores for the PTSD Checklist and the status and treatment progress. The study coordinator (D.Y.) inde- Beck Depression Inventory at these assessment intervals. Separate pendently reviewed chart notes for all providers and treatment Mann-Whitney U tests were then used to compare these change sessions in order to compute clinicians’ self-recorded adherence scores for continuing smokers versus subjects who were abstinent to protocol interventions as a percentage of the total recom- at each assessment interval. mended. This review showed that 80% of behavioral counseling All categorical variables (other than abstinence status) were interventions prescribed by the treatment manual were report- analyzed with chi-square tests. Normally distributed continuous edly administered. variables were analyzed with Student’s t tests. Nonparametric

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1313 PTSD AND QUITTING SMOKING

TABLE 2. Proportion of Subjects Receiving Smoking-Cessation possible) when they were abstinent from smoking (rs= a Medications 0.35, p<0.004, two-tailed). Veterans Veterans Receiving Smoking-Cessation Outcomes Receiving Usual Integrated Standard Subject compliance with follow-up assessments was Care of Care Analysis 83% across all four assessment intervals (83% at month 2, Medication N%N%χ2 (df=1) p 81% at month 4, 81% at month 6, and 84% at month 9). Bupropion 20 60.6 16 48.5 0.55 0.46 Transdermal nicotine 31 93.9 22 66.7 6.13 <0.02 Medical records and/or retrospective subject interview Nicotine gum 29 87.9 14 42.4 13.08 0.001 data were available to determine smoking status for 96% Nicotine spray 1 3.0 3 9.1 0.27 0.61 of the remaining observations. Four percent of the obser- a Prescriptions written for medications filled by the pharmacy. vations were missing and were handled by recording subjects as presumptive smokers. The two study conditions tests (Mann-Whitney U and Spearman’s correlation) were per- did not differ significantly in the number of missing obser- formed on nonnormally distributed variables. vations (p>0.05). Results Figure 1 presents smoking abstinence data for each assessment interval and the results of the generalized esti- Delivery of Study Treatments mating equation analysis. At each assessment interval, the odds of not smoking at that interval were over five times Smoking-cessation medications. Table 2 presents the greater for the subjects in integrated care than the subjects proportion of subjects receiving integrated care and usual in usual standard of care (odds ratio=5.23, 95% confidence standard of care who obtained prescriptions for study interval=1.76–15.54, p<0.002, two-tailed). The proportion medications that were filled by the pharmacy. The subjects randomly assigned to integrated care were significantly of subjects who achieved abstinence at one or more as- more likely than subjects receiving usual standard of care sessment intervals was greater for the subjects in inte- to receive prescriptions for transdermal nicotine and nico- grated care (integrated care=52% versus usual standard of χ2 < tine gum, but the differences between study conditions care=25%) ( =4.82, df=1, p 0.02, two-tailed). The sub- were not significant for other smoking-cessation medica- jects in integrated care were also abstinent at more follow- tions. Subjects in integrated care also received more types up assessment periods (four possible) compared to the of smoking-cessation medications (i.e., gum, patch, spray, subjects in usual standard of care (mean=1.1, SD=1.4, for or bupropion) than subjects in usual standard of care (inte- integrated care versus mean=0.4, SD=0.8, for usual stan- grated care: mean=2.5, SD=0.7; usual standard of care: dard of care) (z=2.38, p<0.02, two-tailed). The repeated 7- mean=1.7, SD=1.0) (z=3.21, p<0.01, two-tailed). Ninety- day point prevalence abstinence rate was 12% for inte- four percent of the subjects in integrated care filled pre- grated care and 3% for the usual standard of care, a statis- scriptions for two or more types of smoking-cessation tically nonsignificant difference (χ2=1.66, df=1, p=0.20, medications over the course of the study compared to 64% two-tailed). of the subjects in usual standard of care (χ2=7.3, df=1, p<0.01, two-tailed). Table 3 shows that subjects in both Quality Assurance Outcomes conditions who received smoking-cessation medications For the group as a whole, scores on the PTSD Checklist were given doses and a course of treatment consistent with and the Beck Depression Inventory gathered at 6 and 9 U.S. Public Health Service practice guidelines (14). months did not change significantly from baseline (all > Behavioral counseling. All subjects randomly assigned p 0.05). Change scores were not significantly different for to integrated care participated in at least one behavioral abstainers versus continued smokers at either assessment > counseling smoking-cessation session, and 88% of the interval (all p 0.05). subjects in usual standard of care had at least one treat- Satisfaction With Treatment ment contact. Subjects in integrated care received an average of 5.15 (SD=1.2) manual-driven protocol smoking- The subjects in integrated care were significantly more cessation sessions, compared to an average of 2.6 (SD=2.1) satisfied with the amount of smoking-cessation treat- sessions for subjects in usual standard of care (z=5.35, ment they received compared to the subjects in usual p<0.0001, two-tailed). Additional provider-initiated fol- standard of care (mean=3.9, SD=0.3, for integrated care low-up contacts (mean=3.9, SD=3.9) were delivered to in- versus mean=3.5, SD=0.7, for the usual standard of care) tegrated care subjects, involving assessment of smoking (z=3.21, p<0.001, two-tailed). Ratings for the quality of status and reapplication of interventions deemed appro- treatment were also significantly higher for the integrated priate by the case manager. The number of smoking-ces- care condition (mean=3.7, SD=0.5, for integrated care sation sessions received by all subjects was significantly versus mean=3.4, SD=0.6, for the usual standard of care) correlated with the number of assessment occasions (four (z=2.08, p<0.04, two-tailed).

1314 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 McFALL, SAXON, THOMPSON, ET AL.

TABLE 3. Quantity of Smoking-Cessation Medications Received by Subjectsa Veterans Receiving Veterans Receiving Integrated Care Usual Standard of Care Analysis Medication and Measure N Mean SD N Mean SD t df p (two-tailed) Bupropion Dose (mg/day) 20 281.3 46.4 16 285.2 33.9 –0.29 33.78 0.77 Days’ supply 20 141.8 114.5 16 97.1 93.2 1.29 33.98 0.21 Number of prescriptions filled 20 3.8 2.8 16 2.9 2.3 1.10 34.00 0.28 Transdermal nicotine Dose (mg/day) 31 23.7 5.4 22 24.0 7.7 –0.12 35.35 0.91 Days’ supply 31 103.0 119.1 22 79.6 58.2 0.95 46.13 0.35 Number of prescriptions filled 31 4.6 4.8 22 3.2 2.4 1.43 46.16 0.16 Nicotine gum Dose (mg/day) 29 7.6 2.8 14 6.2 3.2 1.42 22.70 0.17 Days’ supply 29 79.9 69.5 14 66.1 66.3 0.63 26.92 0.54 Number of prescriptions filled 29 2.8 2.4 14 2.6 2.7 0.26 23.18 0.80 Nicotine sprayb Number of bottles received 1 1.0 3 4.7 3.5 Days’ supply 1 14.0 3 80.0 45.8 Number of prescriptions filled 1 1.0 3 2.0 1.7 a Prescriptions written for medications that were filled by the pharmacy. b Significance testing was not conducted because of the small number of observations in the two study conditions.

Discussion engaging subjects in cessation treatment, as indicated by higher satisfaction ratings for integrated care than usual This study demonstrated the feasibility of training men- standard of care and encouraging medication compli- tal health providers to integrate guideline-based smoking- ance. Additionally, the continuous therapeutic relation- cessation treatment into mental health care for veterans ship afforded by integrated care, as opposed to the epi- with PTSD. PTSD clinic prescribers readily incorporated sodic care provided by the usual standard of care, may the delivery of tobacco-cessation medications into their have improved access to ongoing monitoring and pro- clinical practice. Case managers also documented a high vider-initiated relapse management. Other research has rate of adherence to prescribed behavioral counseling in- shown a “dose-response” relationship between the num- terventions, although the absence of verification through ber of smoking-cessation contacts and smoking outcomes independent ratings of session content limits the validity (14, 33, 35), as well as higher quit rates for extended versus of this finding. The integrated model of smoking-cessa- brief tobacco use treatments (45). In fact, some treatment tion treatment tested here was more effective for PTSD pa- trials suggest that cessation may be related more to the tients than for care provided by VA smoking-cessation number of counseling sessions received than specific specialists, as measured by point-prevalence abstinence. therapeutic methods used for smokers with a positive his- The difference between study conditions on the repeated tory of depression (32–35). Successful smoking-cessation 7-day point prevalence abstinence measure (12% for inte- treatment for veterans with PTSD may similarly require a grated care, 3% for the usual standard of care) was not sig- greater number of contacts over an extended time. nificant, owing to the limited statistical power of the study. Guideline-recommended smoking-cessation treatments Since completing this randomized trial, PTSD clinic staff yielded point prevalence quit rates of 15% to 25% in U.S. have delivered integrated care to 107 additional patients, Public Health Service meta-analyses of 6,000 studies in- indicating sustainability of the intervention in routine volving follow-up assessments of at least 5 months (12). In- clinical practice. (For these patients, carbon monoxide- dividuals with mental disorders were typically excluded ≤ verified [ 10 ppm], repeated 7-day point prevalence absti- from smoking-cessation clinical trials reviewed by the U.S. nence quit rates measured at 4, 6, and 9 months after Public Health Service, as active psychiatric comorbidities treatment onset were 15%.) complicate cessation efforts and are associated with signif- Subjects in integrated care participated in a greater icantly reduced quit rates (5, 46–49). The fact that point number of smoking-cessation counseling sessions than prevalence quit rates observed in this study fell within the subjects in usual standard of care and were significantly range of those reported in meta-analyses of treatment tri- more likely to receive transdermal nicotine and nicotine als is encouraging, given that our subjects were chronically gum. They were also more likely to receive combination mentally ill and experienced moderate to severe symptoms tobacco-cessation pharmacotherapy, which has been of PTSD and depression. Research is clearly needed that shown to improve smoking quit rates over use of mono- improves upon the repeated 7-day point prevalence absti- therapy (14). These findings suggest that integrated care nence quit rate (12%) of the integrated-care condition. Re- was a more effective vehicle than the usual standard of application of integrated care over a longer period in order care for delivering cessation treatments of sufficient in- to promote “recycling” is one possible strategy to test. An- tensity. Integrated care may have been more successful at other is to increase the intensity of initial integrated care

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1315 PTSD AND QUITTING SMOKING

FIGURE 1. Point Prevalence Smoking Abstinence Rates for peated 7-day point prevalence abstinence quit rate of 12% Veterans With Posttraumatic Stress Disorder in Integrated in the present study. Care Versus the Usual Standard of Carea The rate of smoking abstinence for integrated care sub- 50 jects at study termination was less than half that observed Veterans receiving integrated care (N=33) at the 2-month assessment. Several possible mechanisms Veterans receiving standard care (N=33) that may contribute to smoking relapse in PTSD patients 40 deserve further exploration. Specifically, pre- and postcessation negative affective states (24, 32, 33, 44, 46, 62) 30 and increased symptoms of depression immediately following a quit attempt (63) increase relapse susceptibility and may have suppressed sustained quit rates in this study. 20 Heightened withdrawal sensitivity has predicted relapse in smokers with a history of depression in some studies (64, 65) and may possibly explain continued smoking in PTSD 10 patients as well. Percent Abstinent From Smoking Stopping smoking in this treatment trial was not associ- 0 ated with worsening symptoms of PTSD or depression 2469measured at 6 and 9 months after treatment onset. Our re- Time (months) sults are more consistent with studies showing that mood a Significant effect of group (odds ratio=5.23, 95% CI=1.76–15.54, disturbances do not result from stopping smoking (32, 46, p<0.001, two-tailed). 66) than with studies showing that they do (25–27, 67) among individuals with a history of depression. However, interventions in an effort to prevent early relapse to smok- comparing our results with those of others is attenuated ing. Application of integrated care on a systemwide basis, because of the marked differences in study groups (i.e., even with a prolonged smoking quit rate of 12%, still has a current PTSD versus a history of major depression). Three potential for producing 10,000 quitters from within the other possible explanations for this finding also deserve population of PTSD patients seen at the VA alone (esti- consideration. First, transient symptom exacerbations re- mates derived from VA databases [11, 50, 51]). lated to postcessation nicotine withdrawal may have oc- There exist remarkably few smoking-cessation clinical curred shortly after subjects’ cessation attempts but were trials involving individuals with current mental disorders undetected because outcomes were only measured at four that provide a standard of comparison for the present discrete time points in our study. Second, the fact that 86% study. A number of studies have been conducted with of the subjects in the present research were receiving anti- smokers having a history of past major depression (32–36, depressant medications (other than bupropion for smok- 52, 53). However, these studies do not provide the appro- ing cessation) for PTSD may have prevented potential priate standard of comparison for the present study, and deterioration effects. This possibility is supported by the preponderance of evidence from these trials does not evidence showing that sertraline reduces nicotine with- substantiate the hypothesis that history-positive smokers drawal symptoms (53) and that nortriptyline alleviates are less likely to quit than history-negative smokers (54). postcessation negative affect (35) in smokers with a his- Clinical trials involving smokers with current mental dis- tory of major depression. Third, subjects who experienced orders who meet the minimum recommended standard worsening psychiatric symptoms upon trying to quit for abstinence at follow-up assessments (≥6 months [37, smoking may have simply resumed smoking in order to 55]) have focused on schizophrenia and alcohol depen- manage these symptoms. Our data do not support this lat- dence. (Single point prevalence abstinence rates at the ter- ter explanation since symptoms of PTSD and depression minal 6-month follow-up assessment for studies of smok- did not change from baseline to the 6- and 9-month as- ers with schizophrenia: 11% [56], 12% [57], 7%–17% [58], sessment for subjects who continued to smoke compared and 19% [59]; single point prevalence abstinence rates at to those who quit. However, the nonsignificant difference the terminal 12-month assessment for studies of alcohol- between subjects who quit and those who did not on dependent patients: 12% [60] and 13%–19% [61].) Quit symptom measures may reflect low statistical power ow- rates from these studies of clinical samples are roughly ing to the small number of quitters involved in compari- equivalent to, if not slightly lower than, the single-point sons at months 6 and 9. prevalence quit rates for integrated care reported here. Interpretation of this study’s findings may be limited by The only study (61) reporting prolonged abstinence (re- several methodological shortcomings. This investigation peated 7-day point prevalence abstinence at follow-up did not follow the convention of measuring outcomes months 1, 3, 6, and 12) found quit rates of 10% and 12% for from a clearly demarcated quit date or the end of the inter- two active tobacco-cessation treatments involving recov- vention period because smoking-cessation treatment was ering alcoholics. These results are comparable to the re- conceptualized as a continuous process involving reappli-

1316 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 McFALL, SAXON, THOMPSON, ET AL. cation of treatment and multiple quit attempts (31). The References comparability of findings with other research that mea- 1. Narrow WE, Rae DS, Robins LN, Regier DA: Revised prevalence sures outcomes from an initial quit date may therefore be estimates of mental disorders in the United States. Arch Gen limited. Biomarkers of longer-term nonsmoking status Psychiatry 2002; 59:115–123 were not used but should ideally have supplemented ex- 2. Breslau N: Epidemiology of trauma and posttraumatic stress disorder, in Psychological Trauma. Edited by Yehuda R. Wash- pired carbon monoxide in order to verify subject self-re- ington, DC, American Psychiatric Press, 1998, pp 1–29 ports. Continuous abstinence was also not assessed be- 3. Hankin CS, Spiro A III, Miller DR, Kazis L: Mental disorders and cause it is not verifiable with standard biological assays. mental health treatment among US Department of Veterans However, continuous abstinence data based on subject Affairs outpatients: the Veterans Health Study. Am J Psychiatry self-report alone would have been informative because 1999; 156:1924–1930 4. Dobie DJ, Kivlahan DR, Maynard C, Bush KR, Davis TM, Bradley subject reports of smoking status are generally valid (68). KA: Posttraumatic stress disorder in female veterans: associa- Proportions of subjects who received prescriptions filled tion with self-reported health problems and functional impair- by the pharmacy were reported, but a specific measure of ment. Arch Intern Med 2004; 164:394–400 medication adherence was not included. Thus, it is impos- 5. Lasser K, Boyd JW, Woolhandler S, Himmelstein DU, McCormick D, Bor DH: Smoking and mental illness: a population-based sible to determine to what degree, if any, medication ad- prevalence study. JAMA 2000; 284:2606–2610 herence may have affected outcomes. PTSD patients with 6. Cigarette smoking among adults—United States, 2000. MMWR a current substance abuse disorder were included in the Morb Mortal Wkly Rep 2002; 51:642–645 study, but those with current substance dependence were 7. Breslau N, Davis GC, Schultz LR: Posttraumatic stress disorder not. Thus, our findings may not generalize to smokers and the incidence of nicotine, alcohol, and other drug disorders in persons who have experienced trauma. Arch Gen Psy- with PTSD and comorbid substance dependence disor- chiatry 2003; 60:289–294 ders. Finally, a much larger study group is clearly required 8. Beckham JC, Kirby AC, Feldman ME, Hertzberg MA, Moore SD, in order to conclude definitively that integrated care is Crawford AL, Davidson JR, Fairbank JA: Prevalence and corre- more effective than the usual standard of care for smoking lates of heavy smoking in Vietnam veterans with chronic posttraumatic stress disorder. Addict Behav 1997; 22:637–647 cessation in other clinical settings. 9. Beckham JC, Roodman AA, Shipley RH, Hertzberg MA, Cunha This service delivery study was conducted in the spirit GH, Kudler HS, Levin ED, Rose JE, Fairbank JA: Smoking in Viet- of a clinical effectiveness trial in order to promote general- nam combat veterans with post-traumatic stress disorder. J ization to “real world” health care settings (69). To this Trauma Stress 1995; 8:461–472 10. Shalev A, Bleich A, Ursano RJ: Posttraumatic stress disorder: so- end, treatments were represented as they are likely to be matic comorbidity and effort tolerance. Psychosomatics 1990; actually practiced, and broad subject inclusion criteria 31:197–203 were used to ensure that the study group reflected the 11. Miller DR, Kalman D, Ren XS, Lee AF, Niu Z, Kazis L: Health Be- population of VA PTSD clinic patients. Despite threats to haviors of Veterans in the VHA: Tobacco Use: 1999 Large Health Survey of VHA Enrollees. Washington, DC, Department internal validity inherent in such trials (69), this study pro- of Veterans Affairs, Veterans Health Administration, Office of vided a favorable test of the feasibility and outcome of APA Quality and Performance, 2001, p 5 recommendations (49) to implement smoking-cessation 12. Fiore MC, Hatsukami DK, Baker TB: Effective tobacco depen- treatment in practice-based settings for patients with cur- dence treatment. JAMA 2002; 288:1768–1772 rent mental disorders. Additional studies of integrated 13. Beckham JC, Lytle BL, Vrana SR, Hertzberg MA, Feldman ME, Shipley RH: Smoking withdrawal symptoms in response to a models of smoking-cessation treatment for mentally ill trauma-related stressor among Vietnam combat veterans with patients are warranted, given the extraordinarily high posttraumatic stress disorder. Addict Behav 1996; 21:93–101 prevalence of heavy smoking in these individuals (41% [5] 14. US Department of Health and Human Service: Treating To- to 50% [70]). bacco Use and Dependence: Clinical Practice Guideline. Rock- ville, Md, US Public Health Service, 2000 15. Thorndike AN, Rigotti NA, Stafford RS, Singer DE: National pat- Presented in part at the International Society for Traumatic Stress terns in the treatment of smokers by physicians. JAMA 1998; Studies, New Orleans, Dec. 6–9, 2001. From the Northwest Network 279:604–608 Mental Illness Research, Education, and Clinical Center, the Center for 16. Thorndike AN, Stafford RS, Rigotti NA: US physicians’ treatment Excellence in Substance Abuse Treatment and Education, and the Bio- of smoking in outpatients with psychiatric diagnoses. Nicotine statistics Unit of the Health Services Research and Development Center Tob Res 2001; 3:85–91 of Excellence, Department of Veterans Affairs, Seattle, Wash.; and the University of Washington School of Medicine, the Department of Psy- 17. Goldstein MG, Niaura R, Willey-Lessne C, DePue J, Eaton C, Ra- chiatry and Behavioral Sciences, and the School of Nursing, Seattle. kowski W, Dube C: Physicians counseling smokers: a popula- Address correspondence and reprint requests to Dr. McFall, VA Puget tion-based survey of patients’ perceptions of health care pro- Sound Health Care System, PTSD Programs (S-116 MHC), 1660 S. Co- vider-delivered smoking cessation interventions. Arch Intern lumbian Way, Seattle, WA 98108; [email protected] (e-mail). Med 1997; 157:1313–1319 Supported by the Department of Veterans Affairs Northwest Net- 18. Ockene JK: Primary care-based smoking interventions. Nico- work Mental Illness Research, Education, and Clinical Center, the tine Tob Res 1999; 1(suppl 2):S189–S193 Center for Excellence in Substance Abuse Treatment and Education, 19. Thompson RS, Michnich ME, Friedlander L, Gilson B, Grothanus and a grant from the University of Washington Alcohol and Drug LC, Storer B: Effectiveness of smoking cessation interventions Abuse Institute. integrated into primary care practice. Med Care 1988; 26:62– 76

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1317 PTSD AND QUITTING SMOKING

20. Sherman SE, Yano EM, Lanto AB, Simon B, Rubenstein LV: Vet- 38. Velicer WF, Prochaska JO: A comparison of four self-report erans who smoke: do they want to quit and are we helping smoking cessation outcome measures. Addict Behav 2004; 29: them? in Proceedings of the VA Health Services Research and 51–60 Development 19th Annual Meeting. Washington, DC, Depart- 39. Keely JP, Hughes JR, Carpenter MJ: A method to convert pro- ment of Veterans Affairs, 2001, pp 67–68 longed abstinence and point prevalence quit rates, in Proceed- 21. Drake RE, Mercer-McFadden C, Mueser KT, McHugo GJ, Bond ings of the Seventh Annual Meeting of the Society for Research GR: Review of integrated mental health and substance abuse on Nicotine and Tobacco. Middleton, Wis, SRNT, 2001, p 73 treatment for patients with dual disorders. Schizophr Bull 40. Hall SM, Delucchi KL, Velicer WF, Kahler CW, Ranger-Moore J, 1998; 24:589–608 Hedeker D, Tsoh JY, Niaura R: Statistical analysis of randomized 22. Hellerstein DJ, Rosenthal RN, Miner CR: Integrating services for trials in tobacco treatment: longitudinal designs with dichoto- schizophrenia and substance abuse. Psychiatr Q 2001; 72:291– mous outcomes. Nicotine Tob Res 2001; 3:193–202 306 41. Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom K: The 23. Kessler RC, McGonagle KA, Zhao S, Nelson CB, Hughes M, Eshle- Fagerstrom test for nicotine dependence: a revision of the Fag- man S, Wittchen H-U, Kendler KS: Lifetime and 12-month prev- erstrom Tolerance Questionnaire. Br J Addict 1991; 86:1119– alence of DSM-III-R psychiatric disorders in the United States: 1127 results from the National Comorbidity Survey. Arch Gen Psychi- 42. Anda RF, Williamson DF, Escobedo LG, Mast EE, Giovino GA, atry 1994; 51:8–19 Remington PL: Depression and the dynamics of smoking. JAMA 24. Covey LS, Glassman AH, Stetner F: Depression and depressive 1990; 264:1541–1545 symptoms in smoking cessation. Compr Psychiatry 1990; 31: 43. Niaura R, Britt DM, Shadel WG, Goldstein M, Abrams D, Brown 350–354 R: Symptoms of depression and survival experience among 25. Covey LS, Glassman AH, Stetner F: Major depression following three samples of smokers trying to quit. Psychol Addict Behav smoking cessation. Am J Psychiatry 1997; 154:263–265 2001; 15:13–17 26. Glassman AH, Covey LS, Dalack GW, Stetner F, Rivelli SK, Fleiss 44. Cinciripini PM, Wetter DW, Fouladi RT, Blalock JA, Carter BL, Cin- JF, Cooper TB: Smoking cessation, clonidine, and vulnerability ciripini LG, Baile WF: The effects of depressed mood on smok- to nicotine among dependent smokers. Clin Pharmacol Ther ing cessation: mediation by postcessation self-efficacy. J Con- 1993; 54:670–679 sult Clin Psychol 2003; 71:292–301 27. Glassman AH, Covey LS, Stetner F, Rivelli S: Smoking cessation 45. Hall SM, Humfleet GM, Reus VI, Munoz RF, Cullen J: Extended and the course of major depression: a follow-up study. Lancet nortriptyline and psychological treatment for cigarette smok- 2001; 357:1929–1932 ing. Am J Psychiatry 2004; 161:2100–2107 28. Forbes D, Creamer M, Biddle D: The validity of the PTSD Check- 46. Kinnunen T, Doherty K, Militello FS, Garvey AJ: Depression and list as a measure of symptomatic change in combat-related smoking cessation: characteristics of depressed smokers and PTSD. Behav Res Ther 2001; 39:977–986 effects of nicotine replacement. J Consult Clin Psychol 1996; 29. Beck AT, Steer RA, Garbin MG: Psychometric properties of the 64:791–798 Beck Depression Inventory: twenty-five years of evaluation. 47. Rigotti NA: Treatment of tobacco use and dependence. N Engl Clin Psychol Rev 1988; 8:77–100 J Med 2002; 346:506–512 30. Curry SJ, McBride CM: Relapse prevention for smoking cessa- 48. Niaura R, Abrams DB: Smoking cessation: progress, priorities, tion: review and evaluation of concepts and interventions. and prospects. J Consult Clin Psychol 2002; 70:494–509 Annu Rev Public Health 1994; 15:345–366 49. American Psychiatric Association: Practice Guideline for the 31. Abrams DB, Niaura R: Planning evidence-based treatment of Treatment of Patients With Nicotine Dependence. Am J Psychi- tobacco dependence, in The Tobacco Dependence Treatment atry 1996; 153(Oct suppl) Handbook. Edited by Abrams DB, Niaura R, Brown RA, Em- 50. Fontana A, Rosenheck R: Specialized Outpatient PTSD Pro- mons KM, Goldstein MG, Monti PM. New York, Guilford, pp 1– grams (SOPPs) Workload: Fiscal Year 2002. West Haven, Conn, 26 VA Connecticut Health Care System, VHA Northeast Program 32. Brown RA, Kahler CW, Niaura R, Abrams DB, Sales SD, Ramsey Evaluation Center, 2002 SE, Goldstein MG, Burgess ES, Miller IW: Cognitive-behavioral 51. Flora C: Readjustment Counseling Service: Activity Reporting treatment for depression in smoking cessation. J Consult Clin System for FY02. Washington, DC, Department of Veterans Af- Psychol 2001; 69:471–480 fairs, Readjustment Counseling Service, 2002 33. Hall SM, Munoz RF, Reus VI: Cognitive-behavioral intervention 52. Covey LS, Hughes D, Glassman AH, Blazer D, George LK: Ever increases abstinence rates for depressive-history smokers. J smoking, quitting, and psychiatric disorders: evidence from Consult Clin Psychol 1994; 62:141–146 the Durham Epidemiological Catchment Area. Tob Control 34. Hall SM, Munoz RF, Reus VI, Sees KL, Duncan C, Humfleet GL, 1994; 3:222–227 Hartz DT: Mood management and nicotine gum in smoking 53. Covey LS, Glassman AH, Stetner F, Rivelli S, Stage K: A random- treatment: a therapeutic contact and placebo-controlled study. ized trial of sertraline as a cessation aid for smokers with a his- J Consult Clin Psychol 1996; 64:1003–1009 tory of major depression. Am J Psychiatry 2002; 159:1731– 35. Hall SM, Reus VI, Munoz RF, Sees KL, Humfleet G, Hartz DT, 1737 Frederick S, Triffleman E: Nortriptyline and cognitive-behav- 54. Hitsman B, Borrelli B, McChargue DE, Spring B, Niaura R: His- ioral therapy in the treatment of cigarette smoking. Arch Gen tory of depression and smoking cessation outcome: a meta- Psychiatry 1998; 55:683–690 analysis. J Consult Clin Psychol 2003; 71:657–663 36. Hall SM, Humfleet GL, Reus VI, Munoz RF, Hartz DT, Maude-Grif- 55. Pierce JP, Gilpin EA: A minimum 6-month prolonged absti- fin R: Psychological intervention and antidepressant treatment nence should be required for evaluating smoking cessation tri- in smoking cessation. Arch Gen Psychiatry 2002; 59:930–936 als. Nicotine Tob Res 2003; 5:151–153 37. Hughes JR, Keely JP, Niaura R, Ossip-Klein DJ, Richmond RL, 56. Evins AE, Mays VK, Rigotti NA, Tisdale T, Cather C, Goff DC: A pi- Swan GE: Measures of abstinence from tobacco in clinical tri- lot trial of bupropion added to cognitive behavioral therapy als: issues and recommendations. Nicotine Tob Res 2003; 5: for smoking cessation in schizophrenia. Nicotine Tob Res 2001; 13–25; correction, 5:603 3:397–403

1318 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 McFALL, SAXON, THOMPSON, ET AL.

57. Addington J, el Guebaly N, Campbell W, Hodgins DC, Addington 64. Glassman AH, Stetner F, Walsh BT, Raizman PS, Fleiss JL, Cooper D: Smoking cessation treatment for patients with schizophre- TB, Covey LS: Heavy smokers, smoking cessation, and cloni- nia. Am J Psychiatry 1998; 155:974–976 dine. JAMA 1988; 259:2863–2866 58. George TP, Ziedonis DM, Feingold A, Pepper WT, Satterburg CA, 65. Glassman AH, Helzer JE, Covey LS, Cottler LB, Stetner F, Tipp JE, Winkel J, Rounsaville BJ, Kosten TR: Nicotine transdermal patch Johnson J: Smoking, smoking cessation, and major depression. and atypical antipsychotic medications for smoking cessation JAMA 1990; 264:1546–1549 in schizophrenia. Am J Psychiatry 2000; 157:1835–1842 66. Tsoh JY, Humfleet GL, Muñoz RF, Reus VI, Hartz DT, Hall SM: De- 59. George TP, Vessicchio JC, Termine A, Bregartner TA, Feingold A, velopment of major depression after treatment for smoking Rounsaville BJ, Kosten TR: A placebo controlled trial of bupro- cessation. Am J Psychiatry 2000; 157:368–374; correction, 157: pion for smoking cessation in schizophrenia. Biol Psychiatry 1359 2002; 52:53–61 67. Ginsberg D, Hall SM, Reus VI, Munoz RF: Mood and depression 60. Hurt RD, Eberman KM, Croghan IT, Offord KP, Davis LJ, Morse diagnosis in smoking cessation. Exp Clin Psychopharmacol RM, Palmen MA, Bruce BK: Nicotine dependence treatment 1995; 3:389–395 during inpatient treatment for other addictions: a prospective intervention trial. Alcohol Clin Exp Res 1994; 18:867–872 68. Patrick DL, Cheadle A, Thompson DC, Diehr P, Koepsell T, Kinne 61. Burling TA, Burling AS, Latini D: A controlled smoking cessation S: The validity of self-reported smoking: a review and meta- trial for substance-dependent inpatients. J Consult Clin Psychol analysis. Am J Public Health 1994; 84:1086–1093 2001; 69:295–304 69. Glasgow RE, Lichtenstein E, Marcus AC: Why don’t we see more 62. Hall SM, Munoz RF, Reus VI, Sees KL: Nicotine, negative affect translation of health promotion research to practice? rethink- and depression. J Consult Clin Psychol 1993; 61:761–767 ing the efficacy-to-effectiveness transition. Am J Public Health 63. Burgess ES, Brown RA, Kahler CW, Niaura R, Abrams DB, Gold- 2003; 93:1261–1267 stein MG, Miller IW: Patterns of change in depressive symptoms 70. Substance Abuse and Mental Health Services Administration, during smoking cessation: who’s at risk for relapse? J Consult Office of Applied Studies: National Household Survey on Drug Clin Psychol 2002; 70:356–361 Abuse 2000. http://www.DrugAbuseStatistics.samhsa.gov

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1319 Article

Longitudinal Course of Posttraumatic Stress Disorder and Posttraumatic Stress Disorder Symptoms in a Community Sample of Adolescents and Young Adults

Axel Perkonigg, Ph.D. Objective: Few studies have focused on ence new traumatic event(s) during fol- the natural course of posttraumatic stress low-up (odds ratio=5.21, 95% confidence Hildegard Pfister, Dipl.-Inf. disorder (PTSD) and its determinants in interval [CI]=1.95–13.92), to have higher samples of the general population. The rates of avoidant symptoms at baseline Murray B. Stein, M.D., M.P.H. authors examined determinants of remis- (odds ratio=10.16, 95% CI=1.73–59.51), sion and chronicity of PTSD and associa- and to report more help seeking (odds ra- tions with other disorders in a prospective tio=5.50, 95% CI=1.04–29.05), compared Michael Höfler, Dipl.-Stat. community sample. to respondents with remission. Rates of Method: The data were drawn from a incident somatoform disorder (odds ra- Roselind Lieb, Ph.D. prospective, longitudinal epidemiological tio=4.24, 95% CI=1.60–11.19) and other study of adolescents and young adults anxiety disorders (odds ratio=4.07, 95% Andreas Maercker, M.D., Ph.D. (age 14–24 years) in Munich, Germany CI=1.15–14.37) were also significantly as- (N=2,548). The course of PTSD from base- sociated with a chronic course. Hans-Ulrich Wittchen, Ph.D. line to follow-up 34–50 months later was Conclusions: PTSD is often a persistent studied in 125 respondents with DSM-IV and chronic disorder. Specific symptom PTSD or subthreshold PTSD at baseline. clusters—especially avoidant symp- Results: Although 52% of the PTSD cases toms—might be associated with the remitted during the follow-up period, course of PTSD. In addition, the occur- 48% showed no significant remission of rence of new traumatic events differenti- PTSD symptoms. Respondents with a ates PTSD cases with a chronic course chronic course were more likely to experi- from those with remission.

(Am J Psychiatry 2005; 162:1320–1327)

Posttraumatic stress disorder (PTSD) is a prevalent disor- social support, a greater frequency of social phobia, and der for which overall community lifetime prevalence esti- greater avoidant symptoms. mates range from a minimum of 1% in earlier DSM-III stud- Most of our knowledge of the course of PTSD and its de- ies to a maximum of 12.3% in more recent surveys (1–8). terminants is based on prospective (12–15) or retrospec- A few epidemiological studies, most of them retrospective (16) cohort studies involving people who have experi- tive, have focused on the natural course of PTSD and its de- enced specific types of traumatic events (e.g., natural terminants. Chronic DSM-III or DSM-III-R PTSD was fre- disasters). For example, a prospective study among Aus- quently reported in rape victims, victims of torture and tralian firefighters reported a 30% rate of PTSD identified political violence, refugees, and combat veterans (9–11). by assessment with the General Health Questionnaire 29 months after a bush fire (14). Retrospective data from Kilpatrick et al. (9) reported that 16.5% of women who had adult survivors of the Buffalo Creek flood showed that af- been raped continued to meet PTSD diagnostic criteria an ter 14 years, 28% of the survivors with PTSD still had not average of 17 years afterward. Risk factors for lifetime had remission (15). Results from the National Vietnam chronic (duration at least 1 year) PTSD have been reported Veterans Readjustment Study showed that 15% of all male for a random sample of young adults in the Detroit metro- Vietnam veterans still had PTSD 19 years after combat ex- politan area by Breslau and Davis (10). Compared to young posure (16). Some studies found that specific symptom adults with nonchronic PTSD, those with chronic PTSD clusters and clinical features fluctuated over time and had a higher number of DSM-III-R PTSD symptoms and sometimes increased in intensity after several decades of higher rates of interpersonal numbing and overreactivity decline (14). Furthermore, numerous pre-, peri-, and post- to stimuli that symbolized the stressor, as well as higher exposure risk factors have been reported, suggesting a rates of psychiatric comorbidity and other medical condi- multifactorial model of longitudinal course with various tions. Davidson et al. (11) noted significant differences be- pathways (14). For example, studies of acute stress reac- tween chronic and acute DSM-III PTSD in a general popu- tions or acute stress disorder provided support for associ- lation sample; chronic PTSD was associated with reduced ations between the immediate response after a traumatic

1320 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 PERKONIGG, PFISTER, STEIN, ET AL.

TABLE 1. Status at Baseline and 34–50-Month Follow-Up and Course From Baseline to Follow-Up in Respondents With Sub- threshold Posttraumatic Stress Disorder (PTSD) and Full DSM-IV PTSD at Baseline in a Community Sample of Adolescents and Young Adults Age 14–24 Years Respondents With Subthreshold PTSD at Baseline Respondents With Full DSM-IV PTSD at Baseline (N=101)a (N=24) Conditional Conditional Probability Prevalence Probability Prevalence Status and Course N (Weighted %)b (Weighted %) 95% CI N (Weighted %)b (Weighted %) 95% CI Baseline status All respondents (N=2,548) 101 4.4c 3.5–5.4 24 1.3c 0.8–2.0 Trauma victims (N=393) 101 26.1d 21.6–31.3 24 7.6d 5.0–11.5 Male respondents 43 44.8 2.0c 1.4–2.7 4 16.1 0.2c 0.0–0.6 Female respondents 58 55.2 2.4c 1.8–3.2 20 83.9 1.1c 0.7–1.7 Follow-up status (34–50 months) PTSD symptomse 20 21.1 6 25.0 Subthreshold PTSD 20 18.3 2 6.5 Full DSM-IV PTSD 4 4.2 8 25.6 Otherf 57 56.7 8 42.9 Course from baseline to follow-up Remission 57 56.7 8 42.9 Chronic course 44 43.3 16 57.1 Symptomatic (partial remission) 20 21.1 8 31.5 Subthreshold or full DSM-IV PTSD 24 22.2 8 25.6 New traumatic events after baseline 44 44.5 7 22.6 a Respondents with subthreshold PTSD fulfilled the A (traumatic event, fear), B (persistent reexperiences), and E (duration) criteria for DSM-IV PTSD but did not fulfill completely the C (avoidance or numbing of general responsiveness) and/or D (increased arousal) criteria, although they reported at least one symptom in each of the C and D criteria categories with a duration of more than 1 month. b Probability among respondents with subthreshold PTSD or DSM-IV PTSD. c Prevalence in the entire study sample. d Rate among trauma victims. e Respondents with PTSD symptoms fulfilled the A and the B criteria for DSM-IV PTSD but did not meet the criteria for subthreshold PTSD or the full criteria for DSM-IV PTSD. f Symptoms may be present, but they do not fulfill the criteria for PTSD symptoms, subthreshold PTSD, or DSM-IV PTSD (i.e., duration and persistence criteria). event (i.e., acute motor restlessness) and the longitudinal onset of PTSD. We also previously reported the patterns of course of PTSD (17). Some studies suggested a significant incidence of PTSD in this community sample (24). role of cognitive processing style (18), coping behavior In this article, we report results from the follow-up com- (13), and social support (19). However, generalizability of ponent of the Early Developmental Stages of Psychopa- these findings is an unresolved issue. thology study, which focused on the 42-month longitudi- Several epidemiological studies have shown that PTSD nal course of PTSD and PTSD symptoms. We explored the is strongly associated with comorbid disorders (1, 5, 20). following questions: Most of these findings are based on retrospective or cross- sectional data, and the causal explanations of psychiatric 1. How many persons with subthreshold PTSD or full sequelae or temporal associations between PTSD and PTSD at baseline had entire or partial remission of other disorders are unclear (21, 22). The few extant studies PTSD during the follow-up period? on the course of PTSD and psychiatric risk factors showed 2. How do persons with a chronic course differ from that other anxiety disorders and depressive symptoms af- persons who experience remission(s) during follow- ter the onset of PTSD are related to a more chronic course up with respect to specific PTSD characteristics and (12, 23). However, no studies with an epidemiological and risk factors? longitudinal approach have assessed a broader range of 3. Is a chronic course of PTSD associated with the onset psychiatric risk factors after various types of traumatic of other disorders? events in the general population. The high degree of comorbidity was also confirmed in a Method community sample of adolescents and young adults ages 14–24 years in the Early Developmental Stages of Psycho- Sample and Overall Design pathology study (5). We previously reported that 1.3% of The data presented here were collected as part of the Early De- the baseline sample (0.4% of male respondents and 2.2% velopmental Stages of Psychopathology study (25, 26). The Early of female respondents) and approximately 8% of those Developmental Stages of Psychopathology study was designed as who reported traumatic events fulfilled the DSM-IV crite- a prospective-longitudinal survey to explore prevalence and incidence, familial and other risk factors, and comorbidity and ria for PTSD. PTSD and traumatic events were strongly as- course of mental disorders in a population sample of adolescents sociated with other DSM-IV disorders that occurred as ei- and young adults randomly and proportionally drawn from re- ther a primary or secondary disorder before or after the gional registries to represent the distribution of persons age 14–24

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1321 PTSD IN THE COMMUNITY

TABLE 2. Characteristics of Respondents With Posttraumatic Stress Disorder (PTSD) at Baseline and With Complete Remission or a Chronic Course at 34–50-Month Follow-Up in a Community Sample of Adolescents and Young Adults Age 14–24 Years b Respondents With Respondents With Analysis PTSD Remission a Chronic Course of Respondents Without at Follow-Up PTSD at Follow-Up All Respondents New Traumas Characteristics (N=65)a (N=60)a (N=125) (N=74) Weighted Weighted Odds Odds N % N % Ratio 95% CI Ratio 95% CI Gender Male 31 65.1 16 34.9 1.00 1.00 Female 34 46.5 44 53.5 2.24 0.75–6.65 2.47 0.60–10.16 Age group at baseline (years) 14–17 18 57.6 12 42.4 1.00 1.00 18–24 47 53.0 48 47.1 1.52 0.49–4.72 4.42 0.55–35.55 Trauma type Nonassaultive 25 55.4 21 4v4.6 1.00 1.00 Assaultive 40 52.6 39 47.4 1.03 0.39–2.76 0.29 0.06–1.42 Age at onset of baseline trauma (years) ≥12 43 56.1 37 43.9 1.00 1.00 <12 22 48.6 23 51.4 0.69 0.23–2.05 1.10 0.22–5.57 Number of traumas 1 52 57.4 42 42.6 1.00 1.00 >1 13 42.1 18 58.0 1.44 0.52–3.96 1.10 0.21–5.76 Number of symptoms ≤6 26 56.3 17 43.7 1.00 1.00 >6 39 52.4 43 47.6 0.58 0.13–2.63 0.15c 0.02–0.92 Number of avoidant symptoms <2 42 62.2 26 37.8 1.00 1.00 ≥2 23 44.2 34 55.8 1.39 0.40–4.86 10.16c 1.73–59.51 Number of depressive symptoms and interpersonal numbing <2 54 57.5 38 42.5 1.00 1.00 ≥2 11 44.0 22 56.0 1.87 0.61–5.71 1.61 0.14–18.78 Number of arousal symptoms <2 19 59.9 12 40.2 1.00 1.00 ≥2 46 51.8 48 48.2 1.14 0.26–4.92 2.50 0.53–11.81 Impairment due to symptoms Yes, moderate 53 58.7 40 41.3 1.00 1.00 Yes, severe/very severe 12 41.1 20 58.9 1.67 0.49–5.71 1.91 0.23–16.09 Help seeking No 54 59.7 37 40.3 1.00 1.00 Yes 11 38.1 23 61.9 1.47 0.52–4.12 5.50c 1.04–29.05 New traumas between baseline and follow-up No 50 66.9 24 33.1 1.00 Yes 15 33.3 36 66.7 5.21c 1.95–13.92 Presence of any other baseline mental disorderd 37 55.8 34 44.2 0.75 0.29–1.98 0.42 0.10–1.69

Odds Odds Mean Mean Ratio 95% CI Ratio 95% CI

Self-competencee 2.8 3.0 1.57 0.97–2.54 2.31c 1.04–5.11 a Includes respondents with both subthreshold PTSD and full DSM-IV PTSD. Respondents with subthreshold PTSD fulfilled the A, B (traumatic event, fear, and persistent reexperiences), and E criteria (duration) for DSM-IV PTSD but did not fulfill completely the C (avoidance or numbing of general responsiveness) and/or D criteria (increased arousal), although they reported at least one symptom in each of the C and D criteria categories with a duration of more than 1 month. b Multiple logistic regression analyses of variables related to chronic course. c Significant difference between respondents with remission and respondents with a chronic course. d Includes substance use disorders, mood disorders, other anxiety disorders, somatoform disorders, and eating disorders. e Standardized scores are reported; higher scores are associated with lower self-competence. years in the area. After complete description of the study to the completed (response rate=71%). The first follow-up (time 1), subjects, written informed consent was obtained. among the respondents who were age 14–17 years at baseline, The study was divided into three waves. The first wave was con- was conducted an average of 19.7 months (range=15–25.6) after ducted in 1995 (baseline) and included the representative sample the baseline interviews (response rate=88%). The second follow- of persons age 14–24 years (N=3,021). The second wave was con- up (time 2), which was intended to include all baseline partici- ducted in 1996–1997 (time 1) and included only respondents who pants, was conducted in 1998–1999, an average of 42 months were age 14–17 years at baseline (N=1,228). The third wave was (range=34–50) after the baseline interviews (response rate=84%). conducted in 1998–1999 (time 2) and again included all respon- The findings reported here are based on baseline and second fol- dents who were age 14–24 years at baseline (N=2,548). Detailed low-up data (N=2,548). For the younger cohort (respondents age information on the sampling procedure has been reported else- 14–17 years at baseline), information from the first follow-up (i.e., where (25). Briefly, at baseline, a total of 3,021 interviews were on new traumatic events) was added.

1322 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 PERKONIGG, PFISTER, STEIN, ET AL.

The sociodemographic characteristics of the baseline and en- concordance between the Munich-Composite International Diag- tire follow-up (baseline to time 2) sample have been published (5, nostic Interview diagnosis and the clinical diagnosis (29, 30). 26). Briefly, at baseline, most of the respondents were attending For this report, we specified two additional categories: “sub- school (89%) and living with their parents (97.8%). About 10% threshold PTSD” and “PTSD symptoms.” The category of sub- were in job training. The majority was classified as belonging to threshold PTSD refers to persons who fulfilled the A, B (traumatic the middle class (61.4%). Noteworthy changes in sociodemo- event, fear, and persistent reexperiences), and E criteria (dura- graphic characteristics from baseline to the second follow-up tion) for DSM-IV PTSD but did not completely fulfill the C (avoid- were found for school status (42% were attending school at fol- ance or numbing of general responsiveness) and/or D criteria (in- low-up) and employment status (24% were in a job training pro- creased arousal), although they reported at least one symptom of gram and 12% were employed at follow-up). each of the C and D criteria categories with a duration of more than 1 month. The DSM-IV criterion of impairment was not re- Diagnostic Assessment quired (although it was met by some respondents with subthreshold PTSD). Similar subthreshold diagnoses of PTSD have been Diagnostic assessments at the baseline and the two follow-up discussed in the literature (32). The category of “PTSD symptoms” investigations were based on the computer-assisted personal ver- at follow-up refers to persons who fulfilled the A and the B criteria sion of the Munich-Composite International Diagnostic Inter- of DSM-IV PTSD but did not meet the criteria for full or sub- view (27), which allows for the assessment of symptoms, syn- threshold PTSD. The course specifiers of a chronic course or re- dromes, and diagnoses of 48 mental disorders according to the mission refer to the time between the baseline and second follow- DSM-IV criteria and for collection of data on onset, duration, se- up interviews and describe transitions from subthreshold PTSD verity, and psychosocial impairment. Diagnostic findings were to full DSM-IV PTSD, unchanged status between baseline and obtained by using the Munich-Composite International Diagnos- second follow-up, or transitions to remission (not fulfilling the tic Interview/DSM-IV algorithms. At baseline, the lifetime version criteria for the categories defined earlier). Partial remission with of the Munich-Composite International Diagnostic Interview was PTSD symptoms was included in the category of chronic course. used. At each follow-up, the interval version was applied. In all as- Only persons who failed to meet the criteria for DSM-IV PTSD or sessments, the Munich-Composite International Diagnostic In- our operational criteria for subthreshold PTSD or PTSD symp- terview was supplemented by a separate respondents’ booklet toms (as described earlier) were considered to have remission for that included several scales and questionnaires for assessing psy- the purpose of this investigation. chological constructs relevant to the study (26). For the purpose of this examination, we additionally used a self-competence scale Statistical Analysis (28). Test-retest reliability and validity of the full Munich-Com- Data were weighted to consider different sampling probabilities posite International Diagnostic Interview, along with descrip- as well as systematic nonresponse at baseline. The Stata software tions of the Munich-Composite International Diagnostic Inter- package (33) was used to calculate proportions and standard errors view format and coding conventions, have been reported in detail as well as robust confidence intervals (CIs) for weighted data. Mul- elsewhere (29, 30). The test-retest reliability of the diagnostic tiple logistic regression analyses with odds ratios were used to de- modules of the Munich-Composite International Diagnostic In- scribe significant differences between the respondents with remis- terview was fair to good, with kappa values ranging from 0.64 sion and those with a chronic course. Logistic regression analyses (Yule’s Y=0.80) to 0.78 (Yule’s Y=0.82). In tests of validity, a good with adjustment for age and gender were used to calculate odds ra- concordance between clinicians’ diagnoses and interview DSM- tios for incident disorders (i.e., first onset) in the two course speci- IV diagnoses was found for all disorders (kappa=0.50–0.96) except fier groups; the total follow-up sample (without respondents with psychotic disorders (kappa=0.21). PTSD) was used as the reference group. Persons with a preexisting Posttraumatic stress disorder (and other mental disorders) was case (at baseline) of any of the disorders considered in the analysis defined according to the DSM-IV criteria by using the Munich- of incident disorders were excluded from each of these regression Composite International Diagnostic Interview diagnostic algo- analyses because they were not at risk for the incidence of that dis- rithm (27). Details of PTSD diagnosis have been presented previ- order. In addition, to avoid confounding findings with new trau- ously (5). Briefly, a screening question, a written list of 10 groups of matic events, we conducted all logistic regression analyses using potentially traumatic events, and an open-ended question about data from a sample that excluded persons who experienced new any other traumatic events (to avoid the necessity of speaking traumatic events between baseline and follow-up. We did not find about embarrassing and stigmatizing traumas) were presented, selective attrition from baseline to follow-up relevant to our cate- followed by questions about the presence of the DSM-IV A2 crite- gories of DSM-IV PTSD and subthreshold PTSD. rion (intense fear, helplessness, or horror) during each reported event and questions that probed for the most severe (i.e., most distressing to the individual) event, as well as linkages between Results events. If the respondent indicated several qualifying events (ful- Diagnostic and Symptom Status From Baseline filling the DSM-IV A1 and A2 criteria) that did not cluster, only the criteria for the most distressing event were assessed by asking to the Second Follow-Up which of the events had been the most upsetting at the time it oc- As Table 1 shows, at baseline 5.7% (N=125) of the re- curred. For this report, the 10 specific event types plus the open spondents age 14–24 years fulfilled the criteria for either category were grouped into a category with two exclusive groups: assaultive events (horrific experience during war, imprisonment, subthreshold PTSD (4.4%) or full DSM-IV PTSD (1.3%). being taken hostage or kidnapped, physical attacks, sexual abuse, The proportion of female respondents was significantly and rape) and nonassaultive events (serious accidents, experience higher among respondents with full PTSD (83.9%) (odds of natural catastrophes, sudden death or threat of death of associ- ratio=5.22, 95% CI=1.61–16.80). ates, and witnessing traumatic events that happened to others). A total of 56.7% of respondents with subthreshold PTSD Similar categories have been used in other studies (31). One-week test-retest reliability of the Munich-Composite International Di- at baseline and 42.9% of those with full PTSD at baseline agnostic Interview PTSD section was acceptable (kappa=0.79), as reported not meeting the criteria for PTSD symptoms at was the validity (kappa=0.85) assessed by using the diagnostic either the first or the second follow-up assessment, for an

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1323 PTSD IN THE COMMUNITY

TABLE 3. Incident Other Disorders During 34–50-Month Follow-Up Among Respondents With Posttraumatic Stress Disorder (PTSD) at Baseline and With Complete Remission or a Chronic Course in a Community Sample of Adolescents and Young Adults Age 14–24 Yearsa Respondents With PTSD Remission at Follow-Up (N=65) Analysisc Respondents Without All Respondents New Traumas Reference Groupb (N=65) (N=50) Adjusted Adjusted Weighted Weighted Odds Odds Incident Disorders N %N % Ratio 95% CI Ratio 95% CI Any disorder 625 43.6 10 67.3 3.07 0.97–9.69 2.82 0.62–12.88 Substance use disorders 406 16.3 8 19.6 1.60 0.61–4.17 1.97 0.61–6.39 Mood disorders 292 12.8 4 7.8 0.58 0.20–1.70 0.44 0.10–1.93 Anxiety disorders 229 11.5 5 23.3 2.66 0.86–8.26 2.57 0.59–11.22 Somatoform disorders 602 25.7 16 29.6 1.39 0.68–2.84 1.47 0.62–3.47 a Includes respondents with both subthreshold PTSD and full DSM-IV PTSD. Respondents with subthreshold PTSD fulfilled the A (traumatic event, fear), B (persistent reexperiences), and E (duration) criteria for DSM-IV PTSD but did not fulfill completely the C (avoidance or numbing of general responsiveness) and/or D (increased arousal) criteria, although they reported at least one symptom in each of the C and D criteria categories with a duration of more than 1 month. b Includes all respondents in the overall community sample at follow-up who did not have PTSD and did not at baseline fulfill the criteria for the incident disorders under consideration. overall remission rate of 52% (N=65). The probability of 13.92). To avoid confounding between the other predic- having full DSM-IV PTSD at the second follow-up was tors and new traumatic events, we also performed the highest among respondents with baseline full PTSD multiple logistic regression analyses using data from the (25.6%); only 4.2% of those with subthreshold PTSD at subsample of respondents who had not experienced new baseline progressed to full PTSD. The remaining 48% of traumatic events. Significant differences between the two those with full or subthreshold PTSD at baseline reported course specifier groups were found for avoidant symp- experiencing full or subthreshold PTSD or PTSD symptoms (odds ratio=10.16, 95% CI=1.73–59.51), help seeking toms during the follow-up periods and were classified as (odds ratio=5.50, 95% CI=1.04–29.05), and self-compe- having a chronic course (N=60). There were no significant tence (odds ratio=2.31, 95% CI=1.04–29.05). The number differences in the distribution of respondents with sub- of symptoms, however, was negatively associated with threshold PTSD and with full PTSD in the two course spec- chronicity (odds ratio=0.15, 95% CI=0.02–0.92). ifier groups (difference between remission and a chronic Incident Disorders During Follow-Up course among respondents with subthreshold PTSD: odds ratio=0.57, 95% CI=0.21–1.59; difference between remis- Table 3 shows associations between the two course spec- sion and a chronic course among respondents with full ifiers and other incident mental disorders with onset dur- PTSD: odds ratio=1.74, 95% CI=0.62–4.84). However, ing the follow-up period among persons who had never 44.5% of those with subthreshold PTSD and 22.6% of those before fulfilled the criteria for these disorders. We did not with full PTSD experienced new traumatic events during directly compare differences between the two groups, but the follow-up period (odds ratio=2.57, 95% CI=1.64–4.02; we compared rates of onset of new disorders in the two odds ratio=0.89, 95% CI=0.36–2.22, respectively). groups with the rates of incident disorders during follow- up in the total sample of respondents with no PTSD and no Differences Between Complete Remission history of the specific disorder at baseline. In addition, to and a Chronic Course avoid confounding of new traumatic events and the onset Table 2 shows differences in certain baseline PTSD char- of other disorders, we conducted the analysis with data acteristics and additional factors between respondents from a subset of persons who had not experienced new with full remission and those with a chronic course. Over- traumatic events. all, respondents with a chronic course had higher rates of Both course specifier groups experienced other incident most of the PTSD characteristics and other risk factors disorders during the follow-up period, but the overall rate that were assessed, except for assaultive trauma types, was not significantly different in the category “any disorder” number of symptoms, and presence of other baseline from the rate of follow-up disorders in the reference group. mental disorders. However, although we found no significant odds ratios for To identify core variables associated with chronicity, we specific incident disorders among respondents with remis- performed multiple logistic regression analyses with all sion of PTSD at follow-up, a chronic PTSD course was sig- significant variables from Table 2. The experience of new nificantly associated with higher rates of incident somato- traumatic events between baseline and follow-up was the form disorders (odds ratio=4.24, 95% CI=1.60–11.19) and most robust and significant difference between the two incident other anxiety disorders (odds ratio=4.07, 95% CI= course specifier groups (odds ratio=5.21, 95% CI=1.95– 1.15–14.37) during the follow-up period. Table 3 also shows

1324 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 PERKONIGG, PFISTER, STEIN, ET AL.

However, a chronic course of PTSD seems to be significantly associated with incident disorders during follow- up. A higher risk of somatoform disorders or syndromes in Respondents With a Chronic Course of PTSD at Follow-Up (N=60) persons with PTSD has often been described (34) and was Analysisc associated with a chronic course in our study. The lack of Respondents Without All Respondents New Traumas association between mood disorders and chronicity was (N=60) (N=24) somewhat surprising and may be an artifact of the low Adjusted Adjusted number of cases of depression in this young adult sample. Weighted Odds Odds The role of depression in relation to traumatic events and N % Ratio 95% CI Ratio 95% CI PTSD has been discussed in the literature, and further em- 10 62.4 2.45 0.78–7.67 2.45 0.51–11.87 4 9.4 0.92 0.30–2.83 0.44 0.05–3.54 pirical evidence in larger samples is required, especially 11 25.5 2.46 1.12–5.41 2.48 0.75–8.14 with regard to the course of PTSD (22). Although previous 8 25.7 2.77 1.13–6.78 4.07 1.15–14.37 research found associations between substance use disor- 25 56.2 3.79 1.95–7.39 4.24 1.60–11.19 ders and PTSD (5), we could not show that incident sub- c Adjusted odds ratios (with adjustment for age and gender) and 95% CIs from logistic regression analyses comparing rates of inci- stance use disorders were significantly related to course dent disorders in subsets of the follow-up sample (without per- during this relatively brief follow-up period. sons with a preexisting case of any of the disorders under consideration) to the rates in the course specifier groups. Some limitations of the study should be addressed. The findings were restricted to a small number of respondents with full syndromal cases of DSM-IV PTSD at baseline and that new traumatic events were specifically associated with follow-up. But, as we have noted previously, in this repre- incident mood disorders among those with a chronic sentative community sample of adolescents and young course (odds ratio=2.46, 95% CI=1.12–5.41). adults, the prevalence of full-blown DSM-IV PTSD is lower than in other epidemiological studies (5). It is also possible Discussion that our findings from this relatively young, urban Ger- man community sample, which consisted of well-edu- The key findings of the study are as follows: More than cated persons from an area with a relatively high eco- one-half of the sample with full DSM-IV PTSD at baseline nomic status, may not generalize to other populations. remained symptomatic for more than 3 years, and more Our definition of remission refers to the lack of PTSD crite- than one-half fulfilled the criteria for subthreshold PTSD or rion B (reexperiencing) and E (duration), but it is possible DSM-IV PTSD at 34–50-month follow-up. Nearly one-half that some persons might have residual symptoms con- of those with subthreshold PTSD at baseline remained fined to avoidance and/or hyperarousal. Although one symptomatic or fulfilled the criteria for subthreshold PTSD might question whether lack of criterion B and E symp- or DSM-IV PTSD at follow-up. These estimates are close to toms constitutes remission, we felt that continuing to refer those from other epidemiological studies in older popula- to these persons as having PTSD-related symptoms was tions (1, 14). less than parsimonious, because such symptoms might be Compared to our previous findings and findings from attributable to other DSM disorders. Finally, our findings other studies, the predictors of course partly seemed to on course may not generalize to the course after exposure differ from the predictors of traumatic events and predic- to specific types of traumatic events that were rare or did tors of the onset of PTSD. For example, an early onset and not occur at all in the present sample (such as natural ca- number of traumas were not significantly associated with tastrophes or terrorist attacks). the course of PTSD in this sample. However, the experi- In conclusion, the results of this longitudinal study con- ence of new traumas during the follow-up interval distin- firm in a prospective cohort that PTSD is often a persistent guished a chronic course from a more favorable course and chronic disorder. In adolescents and young adults, with remission. Furthermore, a higher number of avoidant exposure to new traumatic events and seeking help for symptoms (from cluster C) at baseline predicted a chronic PTSD symptoms (which may be either an indicator of se- course. Higher self-competence, which may indicate a verity or of coping ability) are associated with poorer out- greater ability to cope with the effects of recurrent or new comes. Avoidant symptoms in particular seem to predict a traumatic events, was associated with a lower risk of chro- chronic course. Efforts to prevent persons from being ex- nicity. It remains to be shown whether these results are posed to new traumatic events during the course of PTSD stable during a longer course of illness or whether they re- could lessen the chronicity of this disorder. Prevention flect fluctuating symptoms (14). It is further noteworthy may be achieved through the implementation of therapies that we found no significant differences for other baseline that include techniques for teaching individuals how to disorders between respondents with remission and those seek safe living environments and nonabusive social and with a chronic course. This finding might be related to the romantic relationships (35). The role of other comorbid high rate of comorbidity of other disorders at baseline in disorders (e.g., anxiety and somatoform disorders) in in- both groups (more than 50%). fluencing the course of PTSD and the possibility that con-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1325 PTSD IN THE COMMUNITY current treatment of these symptom domains might re- 12. Goenjian AK, Steinberg AM, Najarian LM, Fairbanks LA, duce chronicity require further investigation. Tashjian M, Pynoos RS: Prospective study of posttraumatic stress, anxiety, and depressive reactions after earthquake and political violence. Am J Psychiatry 2000; 157:911–916 Received April 23, 2003; revision received June 23, 2004; accepted 13. Asarnow J, Glynn S, Pynoos RS, Nahum J, Guthrie D, Cantwell Aug. 18, 2004. From the Department of Clinical Psychology and Epi- DP, Franklin B: When the earth stops shaking: earthquake se- demiology, Max Planck Institute of Psychiatry; the Department of quelae among children diagnosed for pre-earthquake psycho- Clinical Psychology and Psychotherapy, University of Technology, Dresden, Germany; and the Department of Psychiatry, University of pathology. J Am Acad Child Adolesc Psychiatry 1999; 38:1016– California San Diego, La Jolla, Calif. Address correspondence and re- 1023 print requests to Dr. Perkonigg, Clinical Psychology and Epidemiol- 14. McFarlane AC: Posttraumatic stress disorder: a model of the ogy Department, Max Planck Institute of Psychiatry, Kraepelinstr. 2– longitudinal course and the role of risk factors. J Clin Psychiatry 10, D-80804 Munich, Germany; [email protected] (e-mail). 2000; 61(suppl 5):15–20 This work is part of the Early Developmental Stages of Psychopa- 15. Green BL, Lindy JD, Grace MC, Glesser GC, Leonard AC, Korol M, thology (EDSP) study and is funded by the German Ministry of Re- Winget C: Buffalo Creek survivors in the second decade: stabil- search and Technology (project nos. 01 EB 9405/6, 01 EB 9901/6, 01 ity of stress symptoms. Am J Orthopsychiatry 1990; 60:43–54 EB 140, and 01 EB 0440). Principal investigators are Hans-Ulrich 16. Kulka RA, Schlenger WE, Fairbank JA, Hough RL, Jordan BK, Wittchen, Ph.D., and Roselind Lieb, Ph.D. Current or former staff members of the EDSP group are Kirsten von Sydow, Ph.D., Gabriele Marmar CR, Weiss DS: Trauma and the Vietnam War Genera- Lachner, Dr.rer.biol.hum., Axel Perkonigg, Ph.D., Peter Schuster, tion: Report of Findings From the National Vietnam Veterans Ph.D., Franz Gander, Ph.D., Michael Höfler, Dipl.-Stat., Holger Sonn- Readjustment Study. New York, Brunner/Mazel, 1990 tag, Dipl.-Psych., and Petra Zimmermann, Dr.rer.nat., as well as 17. Harvey AG, Bryant RA: Two-year prospective evaluation of the Esther Beloch, Mag. Phil., Martina Fuetsch, Ph.D., Elzbieta Garczynski, relationship between acute stress disorder and posttraumatic Dipl.-Psych., Alexandra Holly, Dipl.-Psych., Barbara Isensee, Dr. stress disorder following mild traumatic brain injury. Am J Psy- rer.nat., Marianne Mastaler, Ph.D., Chris Nelson, Ph.D., Hildegard Pfis- chiatry 2000; 157:626–628 ter, Dipl.-Inf., Victoria Reed, Ph.D., Andrea Schreier, Dipl.-Psych., Dilek 18. Mayou RA, Ehlers A, Bryant B: Posttraumatic stress disorder af- Türk, Dipl.-Psych., Antonia Vossen, Dipl.-Psych., and Ursula Wunder- ter motor vehicle accidents: 3-year follow-up of a prospective lich, Ph.D. Scientific advisors are Jules Angst, M.D. (Zurich), Jürgen Margraf, Dr.rer.nat. (Basel), Günther Esser, Ph.D. (Potsdam), Kathleen longitudinal study. Behav Res Ther 2002; 30:665–675 Merikangas, Ph.D. (NIMH, Bethesda, Md.), Jim van Os, M.D., Ph.D. 19. Widows MR, Jacobson PB, Fields KK: Relation of psychological (Maastricht), and Ron Kessler, Ph.D. (Harvard University, Boston). vulnerability factors to posttraumatic stress disorder symptomatology in bone marrow transplant recipients. Psychosom Med 2000; 62:873–882 References 20. Maercker A, Schützwohl M: Long-term effects of political imprisonment: a group comparison study. Soc Psychiatry Psychi- 1. Kessler RC, Sonnega A, Bromet E, Hughes M, Nelson CB: Post- atr Epidemiol 1997; 32:434–442 traumatic stress disorder in the National Comorbidity Survey. 21. Breslau N, Davis GC, Peterson EL, Schultz L: Psychiatric se- Arch Gen Psychiatry 1995; 52:1048–1060 quelae of posttraumatic stress disorder in women. Arch Gen 2. Helzer JE, Robins LN, McEvoy L: Posttraumatic stress disorder in Psychiatry 1997; 54:81–87 the general population. N Engl J Med 1987; 317:1630–1634 22. Breslau N, Davis GC, Peterson EL, Schultz LR: A second look at 3. Shore JH, Vollmer WM, Tatum EJ: Community patterns of post- comorbidity in victims of trauma: the posttraumatic stress dis- traumatic stress disorder. J Nerv Ment Dis 1989; 177:681–685 order–major depression connection. Biol Psychiatry 2000; 48: 4. Resnick HS, Kilpatrick DG, Dansky BS, Saunders BE, Best CL: 902–909 Prevalence of civilian trauma and posttraumatic stress disor- 23. Bolton D, O’Ryan D, Udwin O, Boyle S, Yule W: The long-term der in a representative national sample of women. J Consult psychological effects of a disaster experienced in adolescence, Clin Psychol 1993; 61:984–991 II: general psychopathology. J Child Psychol Psychiatry 2000; 4: 513–523 5. Perkonigg A, Kessler RC, Storz S, Wittchen H-U: Traumatic 24. Stein MB, Hoefler M, Perkonigg A, Lieb R, Pfister H, Maercker A, events and posttraumatic stress disorder in the community: Wittchen H-U: Patterns of incidence and psychiatric risk factors prevalence, risk factors and comorbidity. Acta Psychiatr Scand for traumatic events. Int J Methods Psychiatr Res 2002; 11:143– 2000; 101:46–59 153 6. Breslau N, Davis GC, Andreski P, Peterson E: Traumatic events 25. Wittchen H-U, Perkonigg A, Lachner G, Nelson CB: The Early De- and posttraumatic stress disorder in an urban population of velopmental Stages of Psychopathology study (EDSP): objec- young adults. Arch Gen Psychiatry 1991; 48:216–222 tives and design. Eur Addict Res 1998; 4:18–27 7. Breslau N, Kessler RC, Chilcoat HD, Schultz LR, Davis GC, An- 26. Lieb R, Isensee B, von Sydow K, Wittchen H-U: The Early Devel- dreski P: Trauma and posttraumatic stress disorder in the com- opment Stages of Psychopathology study (EDSP): a method- munity. Arch Gen Psychiatry 1998; 55:626–632 ological update. Eur Addict Res 2000; 6:170–182 8. Cuffe SP, Addy CL, Garrison CZ, Waller JL, Jackson KL, McKeown 27.Wittchen H-U, Pfister H (eds): DIA-X-Interviews: Manual für RE, Chilappagari S: Prevalence of PTSD in a community sample Screening-Verfahren und Interview; Interviewheft Längsschnit- of older adolescents. J Am Acad Child Adolesc Psychiatry 1998; tuntersuchung (DIA-X-Lifetime); Ergänzungsheft (DIA-X life- 37:147–154 time); Interviewheft Querschnittuntersuchung (DIA-X 12 9. Kilpatrick DG, Saunders BE, Veronen LJ, Best CL, Von JM: Crimi- Monate); Ergänzungsheft (DIA-X 12 Monate); PC-Programm zur nal victimization: lifetime prevalence, reporting to police, and Durchführung des Interviews (Längs- und Querschnittuntersu- psychological impact. Crime Delinq 1987; 33:479–489 chung); Auswertungsprogramm. Frankfurt, Germany, Swets & 10. Breslau N, Davis GC: Posttraumatic stress disorder in an urban Zeitlinger; 1997 population of young adults: risk factors for chronicity. Am J Psy- 28. Lachner G, Wittchen H-U: Skala zum Vergleich von Kompeten- chiatry 1992; 149:671–675 zen. Munich, Max-Planck-Institut, Eigendruck, 1995 11. Davidson JRT, Hughes D, Blazer DG, George LK: Post-traumatic 29. Reed V, Gander F, Pfister H, Steiger A, Sonntag H, Trenkwalder stress disorder in the community: an epidemiological study. C, Hundt W, Wittchen HU: To what degree does the Composite Psychol Med 1991; 21:731–721 International Diagnostic Interview (CIDI) correctly identify

1326 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 PERKONIGG, PFISTER, STEIN, ET AL.

DSM-IV disorders? testing validity issues in a clinical sample. Int 32. Marshall RD, Olfson M, Hellman F, Blanco C, Guardino M, J Methods Psychiatr Res 1998; 7:142–155 Struening EL: Comorbidity, impairment, and suicidality in sub- 30. Wittchen H-U, Lachner G, Wunderlich U, Pfister H: Test-retest re- threshold PTSD. Am J Psychiatry 2001; 158:1467–1473 33. Stata Reference Manual: Release 6.0. College Station, Tex, Stata liability of the computerized DSM-IV version of the Munich- Corp, 1999 Composite International Diagnostic Interview (M-CIDI). Soc Psy- 34. Andreski P, Chilcoat H, Breslau N: Post-traumatic stress disor- chiatry Psychiatr Epidemiol 1998; 33:568–578 der and somatization symptoms: a prospective study. Psychia- 31. Breslau N, Chilcoat HD, Kessler RC, Davis GC: Previous exposure try Res 1998; 79:131–138 to trauma and PTSD effects of subsequent trauma: results 35. Kubany S, Hill EE, Owens JA, Ianne-Spencer C, McCaig MA, Trem- from the Detroit Area Survey of Trauma. Am J Psychiatry 1999; ayne KJ, Williams PL: Cognitive trauma therapy for battered 156:902–907 women with PTSD (CTT-BW). J Consult Clin Psychol 2004; 72:3–18

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1327 Article

National Trends in Hospitalization of Youth With Intentional Self-Inflicted Injuries

Mark Olfson, M.D., M.P.H. Objective: The authors examined na- (22.1% to 26.9%), antidepressants (10.0% tional trends from 1990 to 2000 in the uti- to 14.0%), and opiates (2.3% to 3.3%) as a Marc J. Gameroff, Ph.D. lization of community hospital inpatient cause of injury, whereas there were de- services by young people (5–20 years of creases in the ingestion of salicylates Steven C. Marcus, Ph.D. age) with intentional self-inflicted injuries. (14.9% to 10.2%) and barbiturates (1.5% Method: Discharge abstracts from a na- to 0.7%). There were significant increases Ted Greenberg, M.P.H. tionally representative sample of commu- in the proportion of subjects with primary nity hospitals were analyzed, with a focus mental disorder discharge diagnoses of on youth discharges (N=10,831) with a di- depressive disorder (29.2% to 46.0%), bi- David Shaffer, M.D. agnosis of intentional self-inflicted injury polar disorder (1.3% to 8.2%), and sub- (ICD-9-CM: E950–E959). Census data were stance use disorder (5.4% to 10.7%) and used to derive national population-based significant decreases in the rate of adjust- rates of self-inflicted injuries requiring in- ment disorders (22.2% to 11.4%) and non- patient treatment. Overall population- mental disorders (31.9% to 13.6%). After based trends in hospitalizations for self-in- excluding cutting, which may be more flicted injury were calculated and stratified closely related to self-mutilation than sui- by gender and age. Among youths hospi- cidal self-injury, the annual hospitaliza- talized with a self-inflicted injury, trends tion rate of youths with self-inflicted inju- were also calculated for length of stay, in- ries declined from 47.2 per 100,000 in patient costs, method of injury, and associ- 1990 to 39.4 per 100,000 in 2000. ated mental disorder diagnoses. Results: The annual hospitalization rate Conclusions: Over the decade of study, of youths with self-inflicted injuries de- young people admitted to community clined from 49.1 per 100,000 in 1990 to hospitals with self-inflicted injuries tended 44.9 per 100,000 in 2000, and the mean to have more severe psychiatric diagnoses length of inpatient stay significantly de- and to be treated during shorter inpatient clined from 3.6 days to 2.7 days. Among stays. These trends suggest that the role of the hospitalized patients, there were in- youth inpatient care has narrowed, be- creases in the rate of cutting (4.3% to coming focused on those with severe psy- 13.2%) and ingestion of acetaminophen chiatric disorders.

(Am J Psychiatry 2005; 162:1328–1335)

Considerable controversy surrounds the proper role whether to admit a young person for inpatient care follow- of short-term inpatient psychiatric treatment for the acute ing a self-inflicted injury (4, 5). care of young people following an intentional self-injury Over the last several years, a variety of cost-containment (1). Whereas some clinical researchers have questioned mechanisms have been developed to encourage the sub- the necessity and usefulness of admitting acutely self-institution of less costly outpatient care for more expensive jurious youth for inpatient psychiatric treatment for other inpatient services. Between 1987 and 1997, the proportion than medical reasons (2), others have developed specific of total national mental health service expenditures de- psychiatric indications for admitting suicidal youth to the voted to inpatient psychiatric treatment declined from hospital (3). In practice, emergency room clinicians must 40.5% in 1987 to 29.6% in 1997 (6). During this period, grapple with determining whether acutely self-injurious many managed care plans developed specific privately children and adolescents can be safely managed from a held criteria for approving inpatient treatment of suicidal medical and psychiatric perspective in treatment settings patients (7). For patients admitted to the hospital, man- that are less restrictive and disruptive than hospital care. A aged care utilization management techniques also sought wide range of clinical considerations—including medical to reduce the length and costs of inpatient care (8, 9). At severity of the attempt, child risk factors, availability of the same time, the number of families covered by man- family support and community resources, cost and reim- aged behavioral health plans increased (8). bursement issues, and concerns over adherence with fol- In the current study, we examine national trends from low-up treatment plans—influence the clinical decision of 1990 to 2000 in community hospital admissions of young

1328 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 OLFSON, GAMEROFF, MARCUS, ET AL. people 5–20 years of age with intentional self-inflicted in- classified into depressive disorders (ICD-9-CM: 296.2, 296.3, juries. Using nationally representative data, we describe 298.0, 300.4, 311), adjustment disorders (309), substance use dis- changes in the rates at which young people were admitted orders (291, 292, 303, 304, 305), personality disorders (301), conduct disorder (312, 313.81), psychotic disorders (295, 297–299), to community hospitals with such injuries and character- bipolar disorder (296.0, 296.1, 296.4–296.9), anxiety disorders ize changes in the method of injury, clinical diagnoses re- (300.0, 300.2, 300.3, 308.3, 309.21, 309.81, 313.0), eating disorders ceived, length of inpatient stay, discharge status, inpatient (307.1, 307.5), attention deficit hyperactivity disorder (314), and costs, and primary payer. As a result of changes in the other mental disorders. A separate category was constructed for managed care environment, we anticipated that there discharges with no mental disorder diagnosis. Because cutting or self-mutilation is rarely associated with would be a decline in the rate of youths hospitalized with completed suicide in young people (12) and tends to be less lethal self-inflicted injuries, a decrease in their length of stay, than other forms of self-injury (13), we also examined overall and an increase in the proportion diagnosed with high- trends in hospitalization of youths with intentional self-injury ex- risk conditions, such as mood and substance use disor- cluding cutting (E956). Finally, total inpatient expenditures were ders, strongly associated with youth suicide. calculated for all youths admitted with self-inflicted injuries in 1990 and 2000 and for all admissions excluding self-injury by cutting. The Consumer Price Index for medical care was used to in- Method flate 1990 to 2000 dollars (14). We used the SUDAAN statistical software package (15) to ac- Data were drawn from the 1990 and 2000 nationwide inpatient commodate the complex sampling design and weights from the samples of the Healthcare Costs and Utilization Project (10). The Healthcare Costs and Utilization Project when calculating means project is sponsored by the Agency for Healthcare Research and and corresponding standard errors and to calculate 95% confi- Quality and includes over 100 clinical and nonclinical variables. dence intervals (CIs) for the rate estimates. The Healthcare Costs and Utilization Project consisted of 6,268,515 computerized discharge reports from a geographically diverse sample of 882 community hospitals in 1990 and 7,450,992 dis- Results charge reports from a diverse sample of 994 hospitals in 2000. Gender and Age Distribution Community hospitals include nonfederal short-term general hospitals and academic medical centers but not specialized psychiat- of Admissions in 2000 ric hospitals. Each year the Healthcare Costs and Utilization In 2000, several gender differences were evident in the Project approximates a 20% stratified sample of U.S. community method of injury, primary mental disorder diagnosis, and hospitals. Selection into the sample is based on a stratified proba- discharge status of youths admitted with self-inflicted in- bility selection of short-stay nonfederal general hospitals. Weights were constructed on the basis of the reciprocal probability of sam- juries (Table 1). Relative to male subjects, female subjects pling to approximate national estimates. All percentages in this re- were significantly more likely to be admitted because of port are weighted to adjust for the sampling probability. harmful ingesting but less likely to be admitted following We limited the analysis to data from youths 5–20 years of age self-inflicted injuries due to cutting, hanging/suffocating, who were admitted to the hospital on an urgent or emergent basis firearms, or gas asphyxiation. Female subjects were also and had a discharge diagnosis for intentional self-inflicted injury (ICD-9-CM: E950.0–E959.9). To estimate population rates of hos- significantly more likely to be discharged with a primary pitalization for self-inflicted injury, population data were culled mental disorder diagnosis of depressive disorder, adjust- from the 1990 and 2000 United States Bureau of the Census (11). ment disorder, or an eating disorder but were less likely to Our first goal was to describe the method of injury, primary be discharged with a diagnosis of substance use disorder, mental disorder diagnosis, length of stay, and discharge status of psychotic disorder, or attention deficit disorder. In addi- youth by gender and age group in 2000. We then described trends tion, female subjects were significantly more likely than in the hospitalization rate of youths with intentional self-injury between 1990 and 2000. We determined rates of intentional self- male subjects to be discharged to home and less likely to injury per 100,000 population, both overall and stratified by age die in the hospital or be discharged to an inpatient facility and gender. Substantial missing race/ethnicity data and changes other than a short-term hospital or skilled nursing or in- in the federal classification of race/ethnicity categories during the termediate care facility. There was no difference in the study period prevented a meaningful analysis of these variables. number of inpatient days in the community hospital be- We then examined trends in the distribution of self-injury method. Injuries were first classified by major category: drug in- tween male (mean=2.9, SE=0.1) and female (mean=2.6, gestion (E950), hanging/suffocating (E953), firearm (E955), gas as- SE=0.1) patients. phyxiation (E951, E952), cutting (E956), and a residual group of The pattern of youth admissions for self-inflicted inju- other types of injury (E954, E957, E958, E959). Psychotropic drug ries in 2000 also varied by patient age (Table 2). As com- ingestion was subsequently subclassified on the basis of discharge pared with younger children (5–14 years of age), the older diagnosis codes: anxiolytics/sedatives (barbiturates, benzodiazepines, and others), antidepressants, antipsychotics, opiates, and youth were significantly less likely to be admitted follow- other/unspecified psychotropic drugs. Nonpsychotropic inges- ing a self-inflicted hanging/suffocating injury. The older tion was subclassified as analgesics (acetaminophen, salicylates, youth were more likely than their younger counterparts to and other/unspecified) and other nonpsychotropic substances. be discharged with a primary mental disorder diagnosis of We then examined trends for mean length of stay, discharge adjustment disorder or a psychotic disorder but less likely status, primary payer, and primary mental disorder diagnoses to be discharged with an attention deficit disorder diagno- among youth admissions with intentional self-injuries. Primary mental disorder was defined as the first-listed diagnosis that was sis or no mental disorder diagnosis. Older youth also a mental disorder (ICD-9-CM: 219–320). Mental disorders were tended to have a shorter length of inpatient stay (mean=

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1329 HOSPITALIZATION TRENDS FOR SELF-INJURING YOUTH

TABLE 1. Clinical Characteristics of Youths With Self- nonpsychotropic drugs. During the same period, signifi- Inflicted Injuries Admitted to Community Hospital Inpatient cant increases were evident in the proportion of hospital- Services in 2000, by Patient Gendera izations associated with ingestions of antidepressant Rate (%) Rate (%) Among Among medications, opiates, and acetaminophen (Table 4). Male Female There was also a significant increase in the proportion of 2 Youths Youths χ hospitalizations that involved cutting and hanging/ Characteristic (N=3,304) (N=7,527) (df=1) p suffocating. Method of injury Ingestion 71.3 87.3 81.7 <0.0001 Mental disorder diagnosis. In both study years, mental < Cutting 17.1 11.5 20.8 0.0001 disorders were the primary discharge diagnosis for most Hanging/suffocating 3.5 0.3 34.6 <0.0001 Firearm 2.7 0.2 25.4 <0.0001 youths hospitalized because of self-injury. Significant in- Gas asphyxiation 1.3 0.1 17.3 <0.0001 creases were specifically observed in the proportion of < Other 5.1 1.9 24.8 0.0001 discharges with depressive disorder, substance use disor- Primary mental disorder diagnosis at discharge der, bipolar disorder, and attention deficit disorder (Table Depressive disorder 39.2 49.0 31.5 <0.0001 5). There was also a significant increase in the proportion Adjustment disorder 8.6 12.6 19.4 <0.0001 of discharges with a personality disorder listed as a sec- Substance use disorder 16.2 8.3 55.6 <0.0001 Personality disorder 0.8 0.9 0.03 0.86 ondary diagnosis. Conduct disorder 1.7 1.5 0.18 0.67 Clinical service characteristics. In both 1990 and 2000, Psychotic disorder 3.7 1.1 23.2 <0.0001 Bipolar disorder 9.9 7.4 6.8 0.009 the majority of youths admitted because of intentional Anxiety disorder 1.0 1.3 1.1 0.29 self-inflicted injuries were discharged to home (Table 5). < Eating disorder 0.0 0.5 23.9 0.0001 Although transfer to short-term hospitals became signifi- Attention deficit disorder 2.3 0.6 16.5 0.0001 cantly less common, transfers to other inpatient facilities Other mental diagnosis 3.0 3.1 0.04 0.83 became more common. Comparatively few young people No mental diagnosis 13.7 13.6 0.02 0.90 died in the hospital or left the hospital against medical ad- Discharge status Home 62.5 69.1 18.8 <0.0001 vice in either year. Also in both years, private insurance Short-term hospital 6.4 6.9 0.5 0.49 was the primary payer in a majority of the hospitaliza- Skilled nursing/ tions. Medicaid, which increased as a percentage of youth intermediate facility 0.9 0.4 3.6 0.06 Other inpatient facility 26.5 22.1 12.8 0.0004 hospitalizations with intentional self-injury, was the sec- Against medical advice 1.9 1.3 2.1 0.15 ond most common primary payer. The mean length of Died in hospital 1.7 0.2 21.9 <0.0001 stay for the hospitalizations significantly declined from 3.6 a Youth defined as 5–20 years of age. Rates and means are nation- days (SE=0.2) in 1990 to 2.7 days (SE=0.1) in 2000 (t=15.8, ally weighted estimates from the Healthcare Costs and Utilization Project national impatient sample for 2000. SUDAAN software was df=10,829, p=0.0001). used to account for the complex survey design. Inpatient expenditures. Total estimated inpatient costs in inflation-adjusted 2000 dollars for youth admissions with 2.7 days, SD=0.1) than younger youth (mean=3.0 days, intentional self-injuries were $167.5 million in 1990 (95% SD=0.2) (t=2.3, df=5990, p=0.02). CI=147.3–187.9) and $168.2 million (95% CI=147.2–187.8) in 2000. After excluding admissions for self-injury due to National 10-Year Hospitalization Trends cutting, the respective estimates were $160.8 million (95% for Self-Injuring Youth CI=141.4–180.2) in 1990 and $149.7 million (95% CI= Rate of hospitalization. The overall rate of community 130.3–169.1) in 2000 in inflation-adjusted 2000 dollars. hospital inpatient admissions due to self-injury among youths did not significantly change between 1990 (49 per Discussion 100,000 youth population) and 2000 (45 per 100,000 youth population) (Table 3). However, the rate of admissions sig- Between 1990 and 2000, there was a statistically nonsignificantly increased for children 5–9 years of age from 0.4 nificant decrease in the annual rate of community hospi- to 2.1 per 100,000 children (z=−3.96, p<0.0001). After ex- tal inpatient service utilization among youths admitted cluding admissions for cutting, the overall rate of inpa- following intentional self-injury (from approximately 49 tient care for youth with intentional self-injuries signifi- per 100,000 in 1990 to 45 per 100,000). This trend roughly cantly decreased from 47.2 per 100,000 youth population parallels the national decline in suicides among youths in 1990 to 39.4 per 100,000 youth population in 2000 (z= 15–19 years of age, from 11.1 per 100,000 in 1990 to 8.2 per 2.83, p=0.0047). 100,000 in 2000 (12). Method of injury. There was a significant decrease in The trend in hospital admissions may portray impor- the proportion of hospitalizations involving drug inges- tant changes in the care of young people who intentionally tion over the 10-year period. Significant decreases were injure themselves. During the period from 1991 to 2001, specifically observed in the proportion of hospitalizations the Youth Risk Behavior Survey reported a substantial in- associated with barbiturates, salicylates, and unspecified crease in the rate of injurious suicide attempts by students

1330 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 OLFSON, GAMEROFF, MARCUS, ET AL.

TABLE 2. Clinical Characteristics of Youths With Self- TABLE 3. Rate of Community Hospital Inpatient Service Inflicted Injuries Admitted to Community Hospital Inpatient Utilization by Youths With Self-Inflicted Injuries in 1990 Services in 2000, by Patient Agea and 2000, by Age and Gendera Rate (%) Rate (%) Hospitalization Rate Among Among (per 100,000 youth population) Patients Patients 1990 2000 5–14 Years 15–20 Years b of Age of Age χ2 Group Mean 95% CI Mean 95% CI Characteristic (N=1,150) (N=4,842) (df=1) p Total (N=10,831) 49.1 44.9–53.3 44.9 40.3–49.5 Method of injury Age (years) 5–9 (N=102) 0.4 0.1–0.6 2.1 1.3–2.9 Ingestion 81.1 82.8 1.2 0.28 10–14 (N=1,873) 29.1 25.1–33.2 25.4 21.6–29.2 Cutting 13.9 13.0 0.5 0.49 15–20 (N=8,856) 105.4 96.4–114.3 97.6 88.0–107.1 Hanging/suffocating 2.4 1.0 6.9 0.009 Gender Firearm 0.7 1.1 1.4 0.23 Male (N=3,304) 29.6 26.7–32.6 26.5 23.4–29.6 Gas asphyxiation 0.4 0.5 0.2 0.63 Female (N=7,527) 69.6 63.1–76.1 64.2 57.6–70.8 Other 3.0 2.9 0.0 0.91 a Primary mental disorder Youth defined as 5–20 years of age. Inpatient treatment is limited diagnosis at discharge to admission type urgent and emergent. Weighted rates were con- Depressive disorder 47.6 45.7 1.0 0.33 structed as follows: for each cell, the numerator was a nationally Adjustment disorder 9.3 11.9 6.0 0.01 weighted estimate from the Healthcare Costs and Utilization Substance use Project national impatient sample from 1990 or 2000. The denom- disorder 7.3 11.5 1.5 0.14 inator for each cell was a U.S. Census population figure from 1990 Personality disorder 0.5 0.9 3.0 0.08 or 2000. Confidence intervals took into account the standard error Conduct disorder 2.3 1.4 3.3 0.07 of the Healthcare Costs and Utilization Project estimate; the U.S. Psychotic disorder 0.7 2.2 17.4 <0.0001 Census data were from a 100% enumeration and do not have stan- Bipolar disorder 7.6 8.3 0.6 0.46 dard errors associated with them. SUDAAN software was used to account for the complex survey design. Anxiety disorder 1.0 1.2 1.1 0.29 b Eating disorder 0.3 0.4 0.1 0.79 Ns are unweighted. Attention deficit disorder 2.3 0.8 8.1 0.005 Other mental havior in young children. While the cause of this increase diagnosis 3.3 3.1 0.1 0.73 remains unknown, some evidence links child exposure to No mental diagnosis 17.9 12.6 14.3 0.0002 video games and movie violence to violent attitudes (24) Discharge status and behavior (25). One small study has suggested that Home 67.8 66.9 0.2 0.68 Short-term hospital 6.1 6.9 0.8 0.37 younger children who attempt suicide report many of the Skilled nursing/ same depressive symptoms common to suicidal adoles- intermediate facility 0.2 0.7 4.8 0.03 cents (26). The recent increase of inpatient admissions of Other inpatient facility 24.1 23.3 0.2 0.63 Against medical young children with intentional self-inflicted injuries high- advice 1.2 1.5 0.7 0.41 lights the importance of prevention and early intervention Died in hospital 0.6 0.7 0.0 0.89 programs that target preschool (27) or early grade school a Youth defined as 5–20 years of age. Rates and means are nationally (28, 29) children at risk for mental health problems. weighted estimates from the Healthcare Costs and Utilization Project national impatient sample for 2000. SUDAAN software was The changing diagnostic profile of young people admit- used to account for the complex survey design. ted to the hospital following intentional self-injury could indicate a more focused approach for especially high-risk in grades 9 through 12, from 1.7% to 2.6% (16, 17). In the youth. Between 1990 and 2000, the proportion of dis- context of increasing base rates of self-injurious behavior, charges in which a mood disorder or substance use disor- a slight decline in inpatient admissions of youths with in- der was the first listed mental disorder significantly intentional self-inflicted injuries suggests that the propor- creased, whereas the proportion of adjustment disorder or tion of suicide attempts resulting in inpatient care has no mental disorder discharges significantly declined. Psy- substantially declined. chological autopsy studies reveal that mood disorders oc- In line with previous epidemiological (18) and clinical cur in approximately two-thirds of youth suicides (30–32) research (19), hospitalization as a result of self-inflicted in- and that substance use disorders occur in up to two-thirds jury was more common among female than male youths. of older boys who complete suicide (30, 32). By contrast, Male youths were less likely than female youths to have a young people with adjustment disorders or no mental depressive disorder (20) or an eating disorder (21) but disorder are considerably less common among youth were more likely to have substance use disorder (22). Male suicides (30, 32). youths were also proportionately more likely than female A trend toward more severe mental disorder diagnoses youths to use more highly lethal methods of self-injury among inpatient youth discharges suggests the importance (23) and to die in the hospital. of developing rapid and efficient diagnostic procedures to For young children ages 5 to 9 years, the rate of hospital- identify young people with high-risk conditions (33). In one ization with self-inflicted injuries, although still compara- recent study, a school-based program for high school stu- tively low, significantly increased during the study period. dents that focused on depression and suicide risk was asso- This trend should alert clinicians to the risks of suicidal be- ciated with a reduction in suicide attempts (34).

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1331 HOSPITALIZATION TRENDS FOR SELF-INJURING YOUTH

TABLE 4. Method of Injury Among Youths With Self- TABLE 5. Community Hospital Inpatient Service Utilization Inflicted Injuries Admitted to Community Hospital Inpatient Characteristics in 1990 and 2000 for Youths With Self- Services in 1990 and 2000a Inflicted Injuriesa Rate (%) Rate (%) Rate (%) Rate (%) Among Among in 1990 in 2000 χ2 Youth Youth Characteristic (N=4,839) (N=5,992) (df=1) p 2 in 1990 in 2000 χ Primary mental diagnosis Method of Injury (N=4,839) (N=5,992) (df=1) p Depressive disorder 29.2 46.0 64.1 <0.0001 Drug ingestion 92.8 82.4 45.0 <0.0001 Adjustment disorder 22.2 11.4 59.7 <0.0001 Psychotropic drug Substance use disorder 5.4 10.7 71.1 <0.0001 ingestion 31.1 33.6 4.2 0.04 Personality disorder 2.0 0.8 8.9 0.003 Anxiolytics/sedatives 9.5 9.2 0.2 0.70 Conduct disorder 1.6 1.6 0.0 0.97 Barbiturates 1.5 0.7 10.3 0.001 Psychotic disorder 1.3 1.9 3.1 0.08 Benzodiazepines 5.5 6.2 1.4 0.24 Bipolar disorder 1.3 8.2 97.4 <0.0001 Other/unspecified 2.5 2.5 0.0 0.83 Anxiety disorder 1.1 1.2 0.1 0.80 Antidepressants 10.0 14.0 31.8 <0.0001 Eating disorder 0.3 0.4 0.2 0.62 Antipsychotics 2.4 2.5 0.2 0.66 Attention deficit disorder 0.1 1.1 36.9 <0.0001 Opiates 2.3 3.3 8.0 0.005 Other mental diagnosis 3.6 3.1 1.0 0.33 Other/unspecified 10.6 12.0 2.8 0.10 No mental diagnosis 31.9 13.6 107.3 <0.0001 Nonpsychotropic drug Personality disorder (not ingestion 69.9 60.2 41.7 <0.0001 necessarily primary) 6.2 8.4 7.7 0.006 Analgesics 44.0 45.4 1.2 0.28 Discharge status < Acetaminophen 22.1 26.9 16.9 0.0001 Home 69.7 67.1 2.0 0.15 < Salicylates 14.9 10.2 28.6 0.0001 Short-term hospital 9.5 6.8 8.5 0.004 Other/unspecified 9.6 11.1 4.6 0.03 Skilled nursing/ < Other substances 36.5 28.4 38.8 0.0001 intermediate facility 1.2 0.6 3.9 0.05 < Cutting 4.3 13.2 39.8 0.0001 Other inpatient facility 15.9 23.4 22.3 <0.0001 Hanging/suffocating 0.6 1.3 10.1 0.002 Against medical advice 3.2 1.5 14.3 0.0002 Firearm 0.9 1.0 0.1 0.71 Died in hospital 0.4 0.6 1.6 0.21 Gas asphyxiation 0.6 0.5 0.1 0.76 Primary payer Other 1.3 2.9 28.2 <0.0001 Private insurance 51.1 59.5 17.2 <0.0001 a Youth defined as 5–20 years of age. Rates are nationally weighted Medicaid 18.8 22.3 5.6 0.02 estimates from the Healthcare Costs and Utilization Project na- Self-pay or no charge 9.2 13.8 18.7 <0.0001 tional impatient sample for 1990 and 2000. Rows are not mutually Medicare 0.5 0.4 0.3 0.59 exclusive because patients may have multiple methods of injury or Other 20.4 4.0 98.4 <0.0001 multiple ingestion types. SUDAAN software was used to account for a the complex survey design. Youth defined as 5–20 years of age. Rates and means are nationally weighted estimates from the Healthcare Costs and Utilization Project national impatient sample for 1990 and 2000. Numbers in The trends in mental disorder diagnoses among youths parentheses are unweighted Ns. There were a number of missing observations in 1990 and 2000 for discharge status (N=13 and 130, hospitalized following self-inflicted injuries may reflect respectively) and primary payer (N=33 for both years). SUDAAN broad changes in diagnostic practices or clinical decision software was used to account for the complex survey design. making in response to managed care restrictions on inpatient care for less severe mental disorders. To explore this and completion (2, 36). In one case/control study of youth possibility, we performed a set of post hoc Healthcare suicide, the odds ratio of suicide completion for bipolar Costs and Utilization Project analyses. The trends ob- disorder approached that for substance use disorders (36). served among admissions with self-inflicted injuries were However, manic or manic-like symptoms in young people also apparent in the larger sample of youth admissions may be difficult to distinguish from symptoms of ADHD with a primary mental disorder diagnosis. For example, (37, 38), and concern exists that there has been a trend to- among youth admissions with primary mental disorder ward overdiagnosis of youth bipolar disorder (39). From diagnoses, the proportion who were diagnosed with an the available data, it is not possible to determine the ex- adjustment disorder declined from 1990 (16.8%) to 2000 tent to which the increase in discharges associated with (6.5%), whereas increases were seen in depressive disorder bipolar disorder represents a true change in diagnostic (25.7% to 34.4%) and bipolar disorder (4.2% to 12.2%). Be- composition as opposed to a change in diagnostic prac- cause similar trends were further observed among youth tices. Little attention has thus far been focused on early in- whose admissions were self-pay or not charged (adjust- tervention in young people with bipolar disorder (40). ment disorder: 19.4% to 7.9%; depressive disorder: 18.4% There was also a trend toward shorter hospitalizations to 31.1%; bipolar disorder: 2.0% to 9.6%), it is unlikely that for young people with self-inflicted injuries. Together with the diagnostic trends are simply a coding response to re- the trend toward limiting care to those with more severe strictive utilization management policies. diagnoses, the shortening of inpatient treatment may be During the decade under study, there was a particularly placing inpatient staff under increased time pressures to impressive increase in the proportion of hospitalizations locate appropriate outpatient care. Under these con- in which bipolar disorder was the leading mental disorder straints, it is perhaps not surprising that an increasing pro- diagnosis. Bipolar disorder in young people has been as- portion of inpatients were transferred to other inpatient sociated with an increased risk of suicide attempts (35) facilities.

1332 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 OLFSON, GAMEROFF, MARCUS, ET AL.

In the acute outpatient management of suicidal young ported rates measure the total number of hospitalizations people, encouraging results have been reported with rather than the individual persons hospitalized. It would rapid-response outpatient psychiatric teams (41), home- be useful to quantify the extent and characteristics of re- based interventions (42, 43), interpersonal problem-solv- peat admissions for attempted suicide. Fifth, stigma asso- ing skills training (44), and brief cognitive behavior ther- ciated with suicide attempts may have resulted in system- apy for family members (45). It is not known whether ac- atic underreporting of intentional self-injury. Last, the cess to these and other relevant outpatient mental health analysis was limited to admissions to short-term commu- services has expanded to help compensate for the narrow- nity hospitals and does not include the large number of ing role of short-term hospitalization in the care of young young people admitted directly to other types of inpatient people with intentional self-inflicted injuries. facilities (55). Several changes occurred in the pattern of intentional Over the last several years, there has been little change self-injury methods. Methods associated with high case in the overall rate of community hospital inpatient hospi- fatality rates, including gas asphyxiation, hanging/suffo- talizations of children and adolescents with intentional cating, and firearms (23), remained relatively uncommon, self-inflicted injuries. However, the proportion of young possibly because of deaths in the community or the emer- self-injurious inpatients with high-risk conditions (in- gency room before hospital admission. Although inges- cluding depressive, bipolar, and substance use disorders) tions declined as a proportion of admissions for inten- has increased, and the length of their inpatient stays has tional self-injury, there were proportionate increases in declined. These changes, especially in light of increasing ingestion of acetaminophen, antidepressants, and opi- injurious youth suicide attempts in the community, indi- ates. These proportionate increases, together with propor- cate that inpatient care may have assumed a narrower and tionate decreases in barbiturate and salicylate ingestions, more limited role in the treatment of suicidal young peo- may reflect changes in the general use of these substances ple. As mental health care professionals have come to rely (46). The increase in acetaminophen ingestion is espe- less extensively on inpatient treatment for the acute man- cially noteworthy because it poses a serious risk of poten- agement of self-injurious young people, community ser- tially fatal hepatic toxicity (47) that may not be appreci- vice needs have likely increased for rapid-response, crisis- ated by young people (48). oriented outpatient care. During the study period, there was also an increase in the inpatient treatment of intentional self-inflicted injury Received Jan. 14, 2004; revision received June 9, 2004; accepted involving cutting. Self-mutilation in young people tends to Aug. 11, 2004. From the New York State Psychiatric Institute/Depart- ment of Psychiatry, College of Physicians and Surgeons of Columbia have a very low potential for lethality (23). In the United University, New York; and the University of Pennsylvania School of States, six adolescents (age range: 15–19 years) committed Social Work, Philadelphia. Address correspondence and reprint re- suicide by injury with a sharp object in 2000 (12). As com- quests to Dr. Olfson, New York State Psychiatric Institute/Department of Psychiatry, College of Physicians and Surgeons of Columbia Univer- pared with other types of self-injury, self-mutilation is clin- sity, 1051 Riverside Dr., New York, NY 10032; [email protected] ically associated with greater patient perceived likelihood (e-mail). of rescue (13), lower perceived certainty of death (13), and lower rates of mood disorders (49). When these presum- References ably less clinically severe discharges were excluded from the analysis, there was a statistically significant decline in 1. Safer DJ: A comparison of studies from the United States and the rate of youths hospitalized with intentional self-injury: Western Europe on psychiatric hospitalization referrals for from 47.2 to 39.4 discharges per 100,000 youth. youths exhibiting suicidal behavior. Ann Clin Psychiatry 1996; 8:161–168 This study has several limitations. First, diagnostic data 2. Brent DA, Kupfer DJ, Bromet EJ, Dew MA: The assessment and are only a crude index of illness severity and risk of subse- treatment of patients at risk for suicide, in American Psychiat- quent suicide. No information was available concerning ric Press Review of Psychiatry, vol 7. Edited by Frances AJ, Hales several known suicide risk factors in young people includ- RE. Washington, DC, American Psychiatric Press, 1988, pp 353– ing prior suicide attempts (50), precipitating stressful life 385 events (51, 52), access to firearms (2), and family psychiat- 3. Stewart SE, Manion IG, Davidson S: Emergency management of the adolescent suicide attempter: a review of the literature. J ric history (51, 53). Second, no independent assessment Adolesc Health 2002; 30:312–325 was available of the clinical diagnostic codes. A growth in 4. Greenhill LI, Waslick B: Management of suicidal behavior in managed care and concepts of medical necessity (54) may children and adolescents. Psychiatr Clin North Am 1997; 20: have resulted in a tendency to code patients as having 641–666 more severe psychiatric disorders to justify admission 5. Morrissey RF, Dicker R, Abikoff H, Alvir JM, DeMarco A, Ko- under precertification utilization management policies. plewicz HS: Hospitalizing the suicidal adolescent: an empirical investigation of decision-making criteria. J Am Acad Child Ado- Third, the data permit no means of distinguishing ad- lesc Psychiatry 1995; 34:902–911 missions primarily for medical care associated with at- 6. Mark TL, Coffey RM, King E, Harwood H, McKusick D, Genaurdi tempted suicide from admissions primarily for treatment J, Dilonardo J, Buck JA: Spending on mental and substance of the underlying psychiatric disturbance. Fourth, the re- abuse treatment, 1987–1997. Health Aff 2000; 19:108–120

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1333 HOSPITALIZATION TRENDS FOR SELF-INJURING YOUTH

7. Hall RC, Platt DE, Hall RCW: Suicide risk assessment: a review of Trial to test the early-starter model of the development of seri- risk factors for suicide in 100 patients who made severe suicide ous conduct problems. Dev Psychopathol 2002; 13:925–943 attempts: evaluation of suicide risk in a time of managed care. 29. Flanagan KS, Bierman KL, Kam CM: Identifying at-risk children Psychosomatics 1999; 40:18–27 at school entry: the usefulness of multibehavioral problem 8. Martin A, Leslie D: Psychiatric inpatient, outpatient, and medi- profiles. J Clin Child Adolesc Psychol 2003; 32:396–407 cation utilization and costs among privately insured youths, 30. Brent DA, Baugher M, Bridge J, Chen T, Chiappetta L: Age- and 1997–2000. Am J Psychiatry 2003; 160:757–764 sex-related risk factors for adolescent suicide. J Am Acad Child 9. Rissmiller DJ, Steer R, Ranieri WF, Rissmiller F, Hogate P: Factors Adolesc Psychiatry 1999; 38:1497–1505 complicating cost containment in the treatment of suicidal pa- 31. Apter A, Bleich A, King RA, Kron S, Fluch A, Kotler M, Cohen DJ: tients. Hosp Community Psychiatry 1994; 45:782–788 Death without warning? a clinical postmortem study of suicide 10. Overview of the HCUP Nationwide Inpatient Sample. Rockville, in 43 Israeli adolescent males. Arch Gen Psychiatry 1993; 50: Md, Agency for Healthcare Research and Quality, May 2002 138–142 11. US Census Bureau: American FactFinder. http://factfinder.cen- 32. Shaffer D, Gould MS, Fisher P, Trautman P, Moreau D, Kleinman sus.gov/ M, Flory M: Psychiatric diagnosis in child and adolescent sui- 12. Centers for Disease Control: Wonder Mortality Data Request cide. Arch Gen Psychiatry 1996; 53:339–348 Screen. http://wonder.cdc.gov/mortICD9J.html 33. Lucas CP, Zhang H, Fisher PW, Shaffer D, Riger DA, Narrow WE, 13. Stanley B, Gameroff MJ, Michalsen V, Mann JJ: Are suicide at- Bourdon K, Dulcan MK, Canino G, Rubio-Stipec M, Lahey BB, tempters who self-mutilate a unique population? Am J Psychi- Friman P: The DISC Predictive Scales (DPS): efficiently screening atry 2001; 158:427–432 for diagnoses. J Am Acad Child Adolesc Psychiatry 2001; 40: 14. United States Department of Labor, Bureau of Labor Statistics, 443–449 Consumer Price Index. http://www.bls.gov/cpi/home.htm 34. Aseltine RH, DeMartino R: An outcome evaluation of the SOS 15. Shah BV, Barnwell BG, Bieler GS: SUDAAN User’s Manual, Re- Suicide Prevention Program. Am J Public Health 2004; 94:446– lease 7.5. Research Triangle Park, NC, Research Triangle Insti- 451 tute, 1997 35. Lewinsohn PM, Seeley JR, Buckley ME, Klein DN: Bipolar disor- 16. Grunbaum JA, Kann L, Kinchen SA, Williams B, Ross JG, Lowry der in adolescence and young adulthood. Child Adolesc Psychi- R, Kolbe L: Youth risk behavior surveillance—United States, atr Clin North Am 2002; 11:461–475 2001. MMWR Surveill Summ 2002; 51:1–62 36. Brent D, Perper J, Moritz G, Allman C, Friend A, Roth C, Schwe- 17. Brener ND, Krug EG, Simon T: Trends in suicide ideation and ers J, Friend A, Roth C, Schweers J, Balach L, Baughm M: Psychi- suicidal behavior among high school students in the United atric risk factors for adolescent suicide: a case-control study. J States, 1991–1997. Suicide Life Threat Behav 2000; 30:304– Am Acad Child Adolesc Psychiatry 1993; 32:521–529 312 37. State RC, Altshuler LL, Frye MA: Mania and attention deficit hy- 18. Lewinsohn PM, Rohde P, Seeley JR, Baldwin CL: Gender differ- peractivity disorder in a prepubertal child: diagnostic and treat- ences in suicide attempts from adolescence to young adult- ment challenges (case conf). Am J Psychiatry 2002; 159:918–925 hood. J Am Acad Child Adolesc Psychiatry 2001; 40:427–434 38. Kim EY, Miklowitz DJ: Childhood mania, attention deficit hyper- 19. Hawton K, Hall S, Simkin B, Bale K, Bond A, Codd S, Stewart A: activity disorder and conduct disorder: a critical review of diag- Deliberate self-harm in adolescents: a study of characteristics nostic dilemmas. Bipolar Disord 2002; 4:215–225 and trends in Oxford, 1990–2000. J Child Psychol Psychiatry 39. Krasa NR, Tolbert HA: Adolescent bipolar disorder: a nine-year 2003; 44:1191–1198 experience. J Affect Disord 1994; 30:175–184 20. Beautrais AL: Suicide and serious suicide attempts in youth: a 40. Conus P, McGorry PD: First-episode mania: a neglected priority multiple-group comparison study. Am J Psychiatry 2003; 160: for early intervention. Aust NZ J Psychiatry 2002; 36:158–172 1093–1099 41. Greenfield B, Larson C, Hechtman L, Rousseau C, Platt R: A 21. Wichstrom L, Rossow I: Explaining the gender difference in self- rapid-response outpatient model for reducing hospitalization reported suicide attempts: a nationally representative study of rates among suicidal adolescents. Psychiatr Serv 2002; 53: Norwegian adolescents. Suicide Life Threat Behav 2002; 32: 1574–1579 101–116 42. Harrington R, Kerfoot M, Dyer E, McNiven F, Gill F, Harrington 22. Rancans E, Alka I, Renberg ES, Jacobsson L: Suicide attempts V, Woodham A, Byford S: Randomized trial of a home-based and serious suicide threats in the city of Riga and resulting con- family intervention for children who have deliberately poi- tacts with medical services. Nord J Psychiatry 2001; 55:279– soned themselves. J Am Acad Child Adolesc Psychiatry 1998; 286 37:512–518 23. Spicer RS, Miller TR: Suicide acts in 8 states: incidence and case 43. Kerfoot M, Harrington R, Dyer E: Brief home-based interven- fatality rates by demographics and method. Am J Public Health tion with young suicide attempters and their families. J Adolesc 2000; 90:1885–1891 1995; 18:557–568 24. Funk JB, Baldacci HB, Pasold T, Baumgardner J: Violence expo- 44. McLeavey B, Daly R, Ludgate J, Murray CM: Interpersonal prob- sure in real-life, video games, television, movies, and the inter- lem-solving skills training in the treatment of self-poisoning pa- net: is there desensitization? J Adolesc 2004; 27:23–39 tients. Suicide Life Threat Behav 1994; 24:382–394 25. Anderson CA: An update on the effects of playing violent video 45. Piacentini J, Rotherman-Borus MJ, Gillis JR, Graae F, Trautman games. J Adolesc 2004; 27:113–122 P, Cantwell C, Garcia-Leeds C, Shaffer D: Demographic predic- 26. Nock MK, Kazdin AE: Examination of affective, cognitive, and be- tors of treatment attendance among adolescent suicide at- havioral factors and suicide-related outcomes in children and tempters. J Consult Clin Psychol 1995; 63:469–473 young adolescents. J Clin Child Adolesc Psychol 2002; 31:48–58 46. Malmstrom M, Sundquist J, Ekedahl A, Johansson SE: Prescrib- 27. Kaminiski RA, Stormshack EA, Good RH, Goodman MR: Preven- ing patterns for analgesics in relation to underprivileged area tion of substance abuse with rural head start children and fam- (UPA) score, mortality, and suicide in 33 municipalities in the ilies: results of project STAR. Psychol Addict Behav 2002; 16(4 providence of Skane, southern Sweden. Soc Psychiatry Psychi- suppl): S11–S26 atr Epidemiol 1999; 344:259–264 28. Bierman KLL, Coie JD, Dodge KA, Greenberg MT, Lochman JE, 47. Schiodt FV, Rochling FA, Casey DL, Lee WM: Acetaminophen McMahon RJ, Pinderhughes EE (Conduct Problems Prevention toxicity in an urban county hospital. N Engl J Med 1997; 337: Research Group): Using the Fast Track Randomized Prevention 1112–1117

1334 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 OLFSON, GAMEROFF, MARCUS, ET AL.

48. Myers WC, Otto TA, Harris E, Diaco D, Moreno A: Acetami- 52. Marttunen MJ, Aro HM, Lonnqvist JK: Precipitant stressors in nophen overdose as a suicidal gesture: a survey of adolescents’ adolescent suicide. J Am Acad Child Adolesc Psychiatry 1993; knowledge of its potential for toxicity. J Am Acad Child Adolesc 32:1178–1183 Psychiatry 1992; 31:686–690 53. Shafii M, Carrigan S, Whittinghill JR, Derrick A: Psychological 49. Fulwiler C, Forbes C, Santangelo SL, Folstein M: Self-mutilation autopsy of completed suicide in children and adolescents. Am and suicide attempt: distinguishing features in prisoners. J Am J Psychiatry 1985; 142:1061–1064 Acad Psychiatry Law 1997; 25:69–77 54. Dana RH, Conner MG, Allen J: Quality of care and cost-contain- 50. Renaud J, Brent DA, Birmaher B, Chiappetta L, Bridge J: Suicide in adolescents with disruptive disorders. J Am Acad Child Ado- ment in managed mental health: policy, education, research, lesc Psychiatry 1999; 38:846–851 advocacy. Psychol Rep 1996; 79:1395–1422 51. Gould MS, Fisher P, Parides M, Flory M, Shaffer D: Psychosocial 55. Manderscheid RW, Henderson MJ (eds): Mental Health, United risk factors of child and adolescent completed suicide. Arch States, 2000: DHHS Publication SMA 02–3537. Washington, DC, Gen Psychiatry 1996; 53:1155–1162 Center for Mental Health Services, 2001

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1335 Article

Racial and Ethnic Differences in Utilization of Mental Health Services Among High-Risk Youths

Ann F. Garland, Ph.D. Objective: Racial and ethnic disparities tice, special education, alcohol and drug in mental health service use have been abuse, and mental health service sec- Anna S. Lau, Ph.D. identified as a major public health prob- tors). Youths and caregivers were inter- lem. However, the extent to which these viewed with established measures of May Yeh, Ph.D. disparities may be accounted for by other mental health service use, psychiatric di- confounding sociodemographic or clini- agnoses, functional impairment, care- cal predictors of service use (e.g., family giver strain, and parental depression. Kristen M. McCabe, Ph.D. income, functional impairment, caregiver strain) is relatively unexplored, especially Results: Significant racial/ethnic group Richard L. Hough, Ph.D. for youth services. The goal of this study differences in likelihood of receiving any was to test for racial/ethnic disparities in mental health service and, specifically, John A. Landsverk, Ph.D. use of a variety of outpatient, inpatient, formal outpatient services were found af- and informal mental health services ter the effects of potentially confounding among high-risk youths, with the effects variables were controlled. Race/ethnicity of other predictive factors controlled. did not exert a significant effect on the use of informal or 24-hour-care services. Method: Participants were 1,256 youths ages 6–18 years who received services in Conclusions: Racial/ethnic disparities a large, publicly funded system of care in service use remain a public health (including the child welfare, juvenile jus- problem.

(Am J Psychiatry 2005; 162:1336–1343)

Children and adolescents significantly underutilize youths. Very little is known about mental health service mental health services, and unmet need for services ap- use rates for high-risk youths in public service sectors pears to be greatest among racial/ethnic minority groups such as child welfare, juvenile justice, and special educa- (1–3). The Surgeon General’s 2001 report on mental health tion services. Youths in these service sectors exhibit high identified racial/ethnic disparities in mental health ser- rates of psychiatric disorders and are at high risk for a vari- vice use as a major public health problem (4). However, ety of other maladaptive outcomes (8–10); racial/ethnic not all studies have found racial/ethnic discrepancies in minority youths tend to be overrepresented in some of mental health service use among youths (5–7). This incon- these service sectors (3, 10). Given that they have already sistency may be related to geographic and/or method- been identified in a service sector, these youths are likely ological differences, such as varying attention to poten- to have substantially higher rates of mental health service tially confounding predictors or variation in the definition use, compared to community samples, but the extent to of mental health services across studies. which there is differential utilization by race/ethnicity for Analyses of nationally representative survey data re- different types of services is not known. Service use dis- vealed lower rates of mental health service use for African parities among high-risk youths would be particularly American and Latino American children, compared to concerning, given the poor prognosis for youths with un- non-Hispanic white children; follow-up analyses account- treated psychopathology (11). ing for potentially confounding variables (e.g., insurance Multiple youth-, family-, and system-level factors have status) indicated that unmet need was greatest among been shown to predict youth mental health service use, Latino American children, compared to non-Hispanic and many of these factors are associated with race/ethnic- white children (2). However, studies in urban areas with ity. For example, one of the most consistent predictors of families of relatively homogeneous socioeconomic status child mental health service use is the caregiver’s percep- found no racial/ethnic differences in service use (6, 7). tion of the strain of caring for the child (12, 13). Recent ev- Likewise, data from a study in a largely rural area revealed idence has identified significant and robust racial/ethnic no difference in specialty or general service use between differences in caregiver strain (14). Contact with the jus- African American and non-Hispanic white children (5). tice system may be a negative predictor of service use (5, All of these studies included community samples of 15), and certain racial/ethnic minority groups are more children and families, as opposed to selected high-risk likely to make contact with the juvenile justice system

1336 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GARLAND, LAU, YEH, ET AL.

(16). In addition, socioeconomic status and insurance other reasons (4%, N=68), and 8% (N=145) were missing child in- coverage are strongly associated with race/ethnicity and terview data because of inability to contact the child or language are predictors of service use (2, 17). Therefore, among differences/difficulties. Two-thirds of the participants were male. The mean age was high-risk youth involved in public service sectors, it is im- 13.7 years (SD=3.3). Most of the parent/caregiver informants portant to examine whether racial/ethnic differences in (hereafter labeled “parents”) were biological parents (72%). Oth- mental health service use are actually attributable to these ers included adoptive or foster parents, stepparents, and a small potentially confounding variables and/or to clinical fac- number of professional caregivers. tors such as psychiatric diagnoses or functional impair- Procedure and Measures ment, which, as expected, are associated with the likelihood of service use (2, 5, 7, 18). After complete description of the study, written informed consent was obtained from the parent and assent was obtained from Most studies of racial/ethnic differences in service use the youths. Parents and youths were interviewed individually have included assessment of some of these potentially con- (usually in their home) about the youth’s mental health use, founding variables, but, to our knowledge, this study is the needs, and a variety of factors potentially associated with mental first to make a comprehensive assessment of these vari- health service use (e.g., caregiver strain, family income). Parents and youths were compensated (up to $40) for their time, which ables in a rigorous test of the robustness of racial/ethnic dif- averaged 3 hours. Interviewer training and reliability checks have ferences, to examine use of a variety of formal and informal been described previously (8, 9). types of mental health service use, and to have adequate The measures used in this study were as follows: representation across four major racial/ethnic groups Services Assessment for Children and Adolescents. Parent (non-Hispanic white, African American, Latino American, and youth versions of the Services Assessment for Children and Asian American/Pacific Islander). The purpose of this study Adolescents (19) were used to assess utilization of different types was to 1) test for racial/ethnic group differences in mental of mental health and substance abuse services. Only past-year health service use among high-risk youths in a public sys- service use was examined in this study. A previous analysis showed that the test-retest reliability of the Services Assessment tem of care and 2) to test whether identified differences per- for Children and Adolescents for past-year service use is excellent sisted when the effects of other factors known to be associ- for parent informants and is good for youth informants age ≥10 ated with service use were taken into account. years (19). In the current study, service use was considered present if it was reported by either the parent or the youth. The Method following types of services were assessed: 1. Outpatient services, including specialty outpatient care Participants (e.g., specialty mental health clinics or private providers), Participants were a subsample of 1,256 of the 1,715 youths in nonspecialty outpatient care (e.g., visit to a pediatrician for the Patterns of Youth Mental Health Care in Public Service Systems emotional/behavioral issues), and outpatient alcohol and study (9). The study was approved by the human subjects protec- drug abuse treatment. tion committees at Children’s Hospital and Health Center in San 2. Twenty-four-hour-care services, including inpatient care in Diego, San Diego State University, and the University of California, a psychiatric hospital or psychiatric unit within a hospital, San Diego. The study participants were randomly selected from an residential treatment center/group home services, and in- enumeration of all youths “active” in one or more of five San Diego patient alcohol and drug abuse treatment. County public sectors of care (alcohol and drug abuse, child wel- 3. Informal services, including self-help groups/peer counseling, fare, juvenile justice, mental health, and public school special ed- counseling from clergy, and services of alternative healers. ucation services) for youths with serious emotional disturbance during the first half of 1997 (total population=12,662). The sample National Institute of Mental Health Diagnostic Interview was selected by simple random sampling techniques and strati- Schedule for Children Version IV. The computer-assisted par- fied by race/ethnicity and restrictiveness of care (care in aggregate ent and youth versions of the National Institute of Mental Health setting versus the home residence). Data were obtained for 67% of Diagnostic Interview Schedule for Children Version IV (20) were the eligible sample in interviews completed between late 1997 and used to assess the presence of past-year DSM-IV psychiatric diag- early 1999. The participants did not differ significantly from the noses. The reliability and validity of the Diagnostic Interview nonparticipants in age, gender, sector affiliation, or racial/ethnic Schedule for Children are well supported (20). To reduce inter- distribution, except that slightly fewer Asian American/Pacific Is- view duration, the mood and anxiety modules were administered landers participated, relative to the eligible sample. only to the youths, because youths are likely the best informants for internalizing disorders (21). The disruptive behavior disorder The 1,256 participants who provided data for the analyses re- module was administered to both parents and youths, and diag- ported here included those with complete diagnostic and service noses were considered present if either respondent’s report met use data from adult and child interviews and those who were in the diagnostic criteria detailed in the Diagnostic Interview Sched- one of the following four largest racial/ethnic groups: non-His- ule for Children scoring algorithms (including diagnosis-specific panic whites (N=554, 44%), Latino Americans (N=332, 26%), Afri- functional impairment). can Americans (N=282, 22%), and Asian American/Pacific Island- ers (N=88, 7%). Race/ethnicity was self-identified for youth ages Composite International Diagnostic Interview substance 11 years and older and was parent-identified for children ages 6– abuse module. The Composite International Diagnostic Inter- 10 years. Eight percent of the total sample of 1,715 (N=135) were view substance abuse module (22) was administered to youths age excluded because they were identified as biracial or in a category 11 years and older. It was used to identify past-year DSM-IV sub- other than the four largest racial/ethnic groups. Nine percent (N= stance use disorders, including abuse and dependence diagnoses 156) were missing adult interview diagnostic data because the infor alcohol, marijuana, hallucinogens, and stimulants. Children terviewee was not English speaking (5%, N=88) or because of younger than age 11 years were assumed to have no substance use

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1337 ETHNICITY AND MENTAL HEALTH CARE

FIGURE 1. Mental Health Service Use Among Youths Age 6– Family income. Total family income was reported by parents 18 Years in a Large, Publicly Funded System of Care by on an incremental scale of annual incomes from ≤$1,000 to Racial/Ethnic Group (N=1,256) ≥$200,000. For these analyses, the distribution was divided into quartiles, as follows: first quartile, ≤$13,000; second quartile, ≤$25,000; third quartile, ≤$45,000; and fourth quartile, >$45,000. 80 Non-Hispanic white (N=554) Latino American (N=332) Insurance status. Parents reported on their insurance coverage 70 African American (N=282) for mental health care. Coverage was categorized as follows: private insurance, government insurance coverage (e.g., MediCal), Asian American/ 60 Pacific Islander (N=88) or no insurance coverage.

50 Data Analysis All analyses were conducted by using Stata (27), with data 40 weighted to represent the enumerated system-of-care population. Chi-square analyses were used initially to test for overall ra- 30 cial/ethnic group differences in rates of service utilization by service category and individual service type. To account for the Percent Who Used Service 20 survey design, the chi-square statistic was converted to an F sta- 10 tistic with noninteger degrees of freedom by using a second-order Rao and Scott’s correction (27, 28). Follow-up logistic regression 0 analyses were used to test for racial/ethnic differences for each Outpatienta 24-hour care Informal Any mental service category after the effects of potential confounding vari- servicesb health servicea ables were controlled. The significance level was set at an alpha of Service Category 0.05, and all tests were two-sided. a Significant difference among racial/ethnic groups (p<0.001, chi- square test). Results b Significant difference among racial/ethnic groups (p<0.02, chi- square test). The youths in this study had high rates of mental health service utilization. Overall, 72% (N=904) had utilized some disorder. The reliability of the Composite International Diagnostic type of mental health service within the past year. The Interview substance abuse module is strong (22). most commonly used services were professional outpa- Children’s Global Assessment Scale. The Children’s Global tient services (64% of the total sample, N=803). Twenty per- Assessment Scale (23) was used to assess global functional im- cent (N=251) of the participants had used informal services pairment. Interviewers assigned ratings on the Children’s Global Assessment Scale after completion of the youth and parent inter- such as self-help groups, and 13% (N=163) had used inpa- views. The standard cutoff of 60 was used to designate clinically tient or residential treatment services. Figure 1 displays the significant functional impairment (24). rates of any reported mental health service use, as well as Columbia Impairment Scale. The Columbia Impairment rates for outpatient, 24-hour-care, and informal services, Scale (25) was used to assess parent- and youth-reported func- for each racial/ethnic group. There were significant differ- tional impairment. The scale has strong psychometric character- ences across racial/ethnic groups for use of any mental istics, and the recommended clinically significant cutoff score of health service (χ2=37.1, df=3, p<0.001), outpatient services 15 was used to identify youths with significant functional impair- χ2 < χ2 ment (25). In this study, youths were considered functionally im- ( =43.0, df=3, p 0.001), and informal services ( =20.1, 2 paired if they met the cutoff criteria for either the Children’s Glo- df=3, p<0.02), but not for 24-hour-care services (χ =7.3, df= bal Assessment Scale or the Columbia Impairment Scale. 3, p=0.23). Non-Hispanic whites had the highest rates of Caregiver Strain Questionnaire. The Caregiver Strain Ques- service use for any mental health service and for outpatient tionnaire (13) was used to assess the parents’ perceptions of the services; Asian American/Pacific Islanders had the lowest burden or impact of caring for a child with behavioral problems. utilization rates for these categories of service. For informal The reliability and validity of this 21-item self-report measure are services use, Latino Americans, Asian American/Pacific Is- well supported (13). Because there is no recommended clinically significant cutoff score, scores were dichotomized for the analy- landers, and non-Hispanic whites had relatively similar ses by splitting at the median. rates, and African Americans had the lowest rate. Center for Epidemiologic Studies Depression Scale. The Table 1 lists the utilization rates by racial/ethnic group for Center for Epidemiologic Studies Depression Scale (26) was used specific types of services within the broad categories re- to assess caregiver depressive symptoms. The 20-item self-report ported earlier. Significant differences across racial/ethnic scale has demonstrated strong reliability (coefficient alphas= groups were found for specialty mental health outpatient 0.85–0.90) and validity for use with diverse populations. services (χ2=60.6, df=3, p<0.001), outpatient alcohol and Police contact. Parents were asked whether the youth had ever drug abuse treatment (χ2=22.1, df=3, p<0.001), inpatient been arrested, picked up by police, or given a warning by police. psychiatric hospital treatment (χ2=10.1, df=3, p<0.02), and This variable was dichotomized to reflect any versus no lifetime χ2 < police contact. use of self-help groups ( =23.1, df=3, p 0.02). There were no significant racial/ethnic group differences for other Parental education. Parents reported on their highest level of education. This variable was also dichotomized to represent nonspecialty outpatient services (e.g., pediatrics/primary those with and without any college-level education. care, emergency room, etc.), residential treatment or group

1338 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GARLAND, LAU, YEH, ET AL.

TABLE 1. Past-Year Mental Health Service Utilization by Youths Age 6–18 Years in a Large, Publicly Funded System of Care by Racial/Ethnic Group (N=1,256)a Weighted Percent Asian Analysis Non-Hispanic Latino African American/ Design- White American American Pacific Islander Based Service (N=554) (N=332) (N=282) (N=88) χ2 df Fp Outpatient services Specialty mental health servicesb 67.6 53.4 51.8 32.8 60.6 3 11.4 <0.001 Other nonspecialty servicesc 20.2 15.6 16.5 19.6 4.6 3 0.97 0.40 Alcohol and drug abuse treatment services 5.4 3.7 1.2 11.3 22.1 3 6.7 <0.001 Any outpatient service 73.5 60.3 57.1 51.5 43.0 3 7.8 <0.001 24-hour-care services Inpatient psychiatric hospital or inpatient psychiatric unit services 6.5 4.6 2.2 3.6 10.1 3 3.7 <0.02 Residential treatment center or group home services 7.4 5.9 7.4 1.9 4.9 3 1.0 0.38 Alcohol or drug abuse treatmentd 3.5 4.4 1.2 2.0 7.3 3 1.2 0.30 Any-24-hour-care service 14.4 14.0 10.2 6.8 7.3 3 1.5 0.23 Informal services Self-help groups, peer counseling 17.5 18.5 7.4 16.7 23.1 3 3.7 <0.02 Counseling from clergy 4.1 5.7 4.4 8.6 4.7 3 1.0 0.39 Alternative healer 1.0 1.9 0.0 1.0 2.7 2 1.7 0.19 Any informal service 20.3 23.5 11.4 21.2 20.1 3 3.5 <0.02 Any mental health servicee 79.0 70.1 63.7 59.3 37.1 3 6.7 <0.001 a Youths were identified from the active services rolls of one or more of five San Diego County public sectors of care (alcohol and drug abuse services, child welfare services, juvenile justice services, mental health services, and public school special education services for youths with serious emotional disturbance) during the first half of 1997. b Includes visits to professional psychologists, counselors, community mental health clinics, and partial hospitalization or day treatment programs. c Includes visits to pediatricians and physicians and in-home therapy or emergency room visits for emotional or behavioral reasons. d Includes treatment in an inpatient setting or residential treatment center for substance abuse problems. e Includes all outpatient, 24-hour-care, and informal service types listed in the table. home services, inpatient alcohol and drug abuse treatment, Of secondary interest, many of the potentially con- counseling from clergy, or services of alternative healers. founding variables were associated with likelihood of ser- Table 2 presents the racial/ethnic group characteristics vice use. For example, use of any mental health service that were potentially associated with mental health ser- was positively associated with female gender, higher care- vice use. Given the presence of significant racial/ethnic giver strain, contact with alcohol and drug abuse services group effects on some of these variables, logistic regres- or mental health services in the year before the study, a sion analyses were used to test whether the racial/ethnic DSM-IV non-substance-use diagnosis, and functional im- differences in service use identified earlier persisted after pairment. Use of any mental health service was less likely the effects of these variables were controlled. for youths with families in the middle-range incomes, compared to the highest range. Table 3 presents the results of the regression analyses. The racial/ethnic group variables added significantly to the prediction of likelihood of use of outpatient services and Discussion any mental health service, after the effects of all other po- The results of this study indicate that there are significant tentially confounding variables were taken into account. racial/ethnic differences in parent- and youth-reported uti- Specifically, the last column indicates that African Ameri- lization of some types of mental health services among cans and Asian American/Pacific Islanders were approxi- high-risk youths and that these differences persist when the mately one-half as likely to receive any mental health effects of other significant predictors of service use are taken service, compared to non-Hispanic whites (African Ameri- into account. Previous research has identified racial/ethnic cans: odds ratio=0.56, 95% confidence interval [CI]= disparities in service use, but to our knowledge no other 0.37–0.84; Asian American/Pacific Islanders: odds ratio= study has tested the robustness of these differences as rigor- 0.44, 95% CI=0.25–0.77). Outpatient service use was also ously by including as comprehensive an array of potentially approximately one-half as likely for African Americans confounding variables. The study also extends the literature (odds ratio=0.54, 95% CI=0.36–0.80) and Asian American/ by analyzing the use of multiple types of mental health ser- Pacific Islanders (odds ratio=0.48, 95% CI=0.27–0.85) than vices and by revealing that racial/ethnic differences in ser- for non-Hispanic whites. Race/ethnicity did not contribute vice use are not consistent across all service types. to the likelihood of use of 24-hour-care or informal services As expected, this group of high-risk youths identified after the effects of other potentially confounding variables from public service sectors (child welfare, juvenile justice, were controlled. alcohol and drug abuse, mental health, and public school

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1339 ETHNICITY AND MENTAL HEALTH CARE

TABLE 2. Sociodemographic and Clinical Characteristics of Youths Age 6–18 Years in a Large, Publicly Funded System of Care by Racial/Ethnic Group (N=1,256)a Weighted Percent Asian Analysis Non-Hispanic Latino African American/ White American American Pacific Islander Design- Characteristic (N=554) (N=332) (N=282) (N=88) χ2 df Based F p Age (years) 86.5 6 8.7 <0.001 6–11 37.1 18.7 30.4 6.2 12–15 28.3 30.3 32.7 27.6 16–18 34.6 51.0 32.9 66.1 Sex 7.1 3 1.4 0.24 Female 32.6 33.5 40.0 29.4 Male 67.4 66.5 60.0 70.6 Family income by quartile First (≤$13,000/year) 24.9 44.0 27.8 28.3 92.5 9 5.8 <0.001 Second (≤$25,000/year) 23.0 25.5 35.6 30.5 Third (≤$45,000/year) 24.3 18.4 21.8 20.4 Fourth (>$45,000/year) 27.8 12.1 14.8 20.7 Any college education 41.6 10.1 29.9 22.7 132.4 3 28.8 <0.001 Service sector affiliation 12.9 3 2.5 0.06 Alcohol and drug abuse services, mental health services, or public school–based services 68.2 59.5 59.0 65.4 Child welfare services or juvenile justice services 31.8 40.5 41.0 34.6 Diagnosis Any DSM-IV diagnosis 62.7 48.3 48.5 45.5 31.9 3 6.3 <0.001 Any Composite International Diagnostic Interview substance abuse module diagnosis 15.5 20.6 10.8 23.4 16.0 3 2.9 <0.04 Functional impairment by either Columbia Impairment Scale or Children’s Global Assessment Scale cutoffs 67.5 63.2 63.7 56.8 5.7 3 1.1 0.35 Caregiver strain score above median 49.1 40.6 33.3 44.8 23.4 3 4.6 <0.001 Police contact, lifetime 41.7 52.5 42.8 60.2 21.6 3 4.1 0.004 Caregiver depression score above clinical cutoff point (Center for Epidemiologic Studies Depression Scale) 28.8 32.2 30.6 41.9 7.1 3 1.4 0.24 a Youths were identified from the active services rolls of one or more of five San Diego County public sectors of care (alcohol and drug abuse services, child welfare services, juvenile justice services, mental health services, and public school special education services for youths with serious emotional disturbance) during the first half of 1997. special education services for youths with serious emo- Americans and Asian American/Pacific Islanders were less tional disturbance) had very high rates of mental health likely to use outpatient services, compared to non-His- service use. Nearly three-quarters of the participants used panic whites. Latino American youths also had lower rates some mental health service within the past year, com- of outpatient service use, compared to non-Hispanic white pared to utilization estimates for community samples youths, but the statistical significance of this effect was re- ranging from 6% to 21% (2, 24). High utilization rates duced by the inclusion of other predictor variables. Within would be expected in this high-risk sample of youths iden- the broad category of outpatient services, there were sig- tified from public service sectors. However, there were significant bivariate racial/ethnic group differences in utiliza- nificantly different rates of service use across racial/ethnic tion rates for specialty mental health and alcohol and drug groups; 79% of non-Hispanic white youths received a abuse services but not for other nonspecialty care, which mental health service, compared to 59% of Asian Ameri- included pediatric visits. Previous research has also found can/Pacific Islanders, 64% of African Americans, and 70% no racial/ethnic difference in receipt of mental health ser- of Latino Americans. After the effects of potentially con- vices in primary care pediatric offices (29). founding variables were controlled, the youths’ race/ethnicity was still a significant predictor of any service use, Thirteen percent of the participants used some 24- and African American and Asian American/Pacific Is- hour-care service in the past year, including psychiatric lander youths were approximately one-half as likely as hospitalization, alcohol and drug abuse treatment, or res- non-Hispanic white youths to receive any service. idential treatment. There was no overall effect of race/eth- Significant racial/ethnic differences were found for only nicity on likelihood of use of this category of services, ei- certain types of services. Outpatient services were used ther before or after the effects of other predictors were most frequently overall, and there was a significant racial/ controlled. There was a significant bivariate group differ- ethnic group difference for this category of service use. Af- ence in use of psychiatric hospitalization, with non-His- ter the effects of other predictors were controlled, African panic whites reporting the highest rate of use, but this ef-

1340 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GARLAND, LAU, YEH, ET AL.

TABLE 3. Association of Mental Health Service Use With Model Variables Among Youths Age 6–18 Years in a Large, Publicly Funded System of Care (N=1,256)a Model Variable Outpatient Services 24-Hour-Care Services Informal Services Any Mental Health Service Odds Ratio CI Odds Ratio CI Odds Ratio CI Odds Ratio CI Potentially confounding variables Demographic characteristics Age 0.98 0.93–1.0 1.1 0.98–1.2 1.2*** 1.1–1.3 1.0 0.97–1.1 Female sexb 1.7** 1.2–2.4 1.6* 1.1–2.5 0.88 0.57–1.4 1.5* 1.1–2.2 Family/social factors Caregiver strain 1.8** 1.2–2.7 2.8*** 1.6–5.0 1.9* 1.1–3.3 2.2** 1.4–3.5 Caregiver depression 1.1 0.74–1.5 0.86 0.54–1.4 0.92 0.58–1.5 1.0 0.68–1.5 Police contact 0.54** 0.37–0.79 1.3 0.81–2.1 2.7*** 1.6–4.3 0.60* 0.40–0.90 Recruitment from alcohol and drug abuse services, mental health services, or public school–based services 1.4 0.99–1.9 7.4*** 3.3–16.6 1.1 0.67–1.7 1.5* 1.1–2.2 Family resources Incomec First quartile (≤$13,000/year) 0.69 0.42–1.1 0.70 0.38–1.3 1.3 0.67–2.4 0.68 0.40–1.2 Second quartile (≤$25,000/year) 0.51** 0.32–0.82 0.57 0.29–1.1 0.90 0.47–1.7 0.52* 0.31–0.87 Third quartile (≤$45,000/year) 0.50** 0.31–0.83 0.52* 0.27–0.97 1.1 0.58–1.9 0.45** 0.26–0.75 Any college education 1.2 0.81–1.7 0.77 0.48–1.2 0.94 0.58–1.5 1.1 0.73–1.6 Insuranced None 0.89 0.43–1.8 0.59 0.23–1.5 0.53 0.23–1.2 0.82 0.36–1.9 Government (e.g., MediCal) 1.1 0.72–1.6 0.58* 0.34–0.99 0.67 0.42–1.1 0.86 0.57–1.3 Diagnosis and impairment Any Diagnostic Interview Schedule for Children diagnosis 1.7** 1.1–2.5 0.88 0.48–1.6 1.2 0.74–2.0 1.6* 1.1–2.5 Any Composite International Diagnostic Interview substance abuse module diagnosis 0.90 0.54–1.5 1.3 0.73–2.4 1.6 0.99–2.7 1.0 0.56–1.8 Functional impairment 1.4 0.95–2.0 1.2 0.61–2.5 1.5 0.88–2.7 1.5* 1.0–2.2

F p F p F p F p Logistic regression analysis for potentially confounding variables 6.8 <0.0001 5.3 <0.0001 8.1 <0.0001 6.8 <0.0001

Odds Ratio CI Odds Ratio CI Odds Ratio CI Odds Ratio CI Race/ethnic groupe Latino American 0.71 0.46–1.1 1.0 0.57–1.8 1.1 0.66–1.7 0.77 0.48–1.2 African American 0.54** 0.36–0.80 1.1 0.61–1.9 0.55* 0.30–0.99 0.56** 0.37–0.84 Asian American/Pacific Islander 0.48* 0.27–0.85 0.40* 0.17–0.96 0.70 0.36–1.4 0.44** 0.25–0.77

F p F p F p F p Logistic regression analysis for race/ ethnic group variable 4.2 0.0006 1.6 0.19 1.9 0.14 4.1 0.007 Logistic regression analysis for all variables 6.3 <0.001 5.0 <0.001 7.2 <0.001 5.9 <0.001 a Youths were identified from the active services rolls of one or more of five San Diego County public sectors of care (alcohol and drug abuse services, child welfare services, juvenile justice services, mental health services, and public school special education services for youths with serious emotional disturbance) during the first half of 1997. b Reference group: male participants. c Reference group: families with income in the fourth quartile (>$45,000/year). d Reference group: participants with private insurance. e Reference group: non-Hispanic white participants. *<0.05. **p<0.01. ***p<0.001. fect was not statistically significant when the effects of use these services significantly more frequently than did other predictor variables were controlled. non-Hispanic whites, there was no evidence that higher Approximately 20% of the participants used informal ser- rates of informal service use among minority groups com- vices, including self-help or peer counseling groups, coun- pensated directly for lower rates of use of professional ser- seling from clergy, and alternative healers. There was a vices. Previous research with adults in the community sug- significant racial/ethnic group difference in use of this cate- gested that Latino Americans with mental health problems gory of services, with Latino American youths exhibiting often utilize informal or nonspecialty services (30). the highest rates of use and African American youths the The findings regarding the other (i.e., nonrace/ethnicity) lowest. However, after the effects of other predictors were potentially confounding predictors of any mental health controlled, there was no racial/ethnic group difference. service use were generally consistent with previous find- Given that the racial/ethnic minority participants did not ings for youths in the community. Positive predictors in-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1341 ETHNICITY AND MENTAL HEALTH CARE cluded diagnosis and impairment, caregiver strain, female some contact with the public system of care, and thus the gender, and involvement with the alcohol or drug abuse findings do not generalize to general community samples. treatment, mental health, or school-based service sectors; The present study was limited to a cross-sectional anal- negative predictors included moderate family income and ysis of self-reported mental health service use and could police contact (5, 17, 31). The only factor associated with not elucidate the causes of the racial/ethnic disparities all types of service use was parental report of caregiver that were identified. More research is needed to examine strain. Previous research has supported the importance of the dynamic and complex processes that lead to youth this factor in youth mental health service use (7, 12, 13, 31), mental health service utilization and the specific reasons and significant racial/ethnic differences in caregiver strain for these racial/ethnic disparities. The network/episode have been identified, with non-Hispanic whites generally model of access to care, adapted for children and adoles- reporting higher levels of strain, compared to other groups cents, suggests that utilization of services is affected by (14). These results confirm the strength of the association multiple interacting factors, ranging from individual between caregiver strain and service use but indicate that (child and family) help-seeking preferences to broader racial/ethnic differences in reported strain do not entirely system-level factors, including access, availability, referral explain differences in service utilization. practices, and funding policies (33). In terms of individual Contact with the police and/or juvenile justice system characteristics, culturally influenced cognitive explana- was associated with a significantly lower likelihood of pro- tory models for constructs such as problem recognition, fessional outpatient service use but a higher likelihood of etiological attribution, and relevance of mental health ser- informal service use. The high prevalence of substance vices have been hypothesized to help explain racial/ethnic abuse problems among these youths is likely to be associ- differences in service use in adult populations (34, 35). In support of this theory for youth populations, racial/ethnic ated with referral to Alcoholics Anonymous–related ser- differences in parental beliefs about children’s behavioral vices (8, 18). Referral rates from the justice system to for- problems have been identified (36, 37), and these differ- mal mental health services are much lower than would be ences may partly explain service use patterns. expected, given the high level of need for mental health services in this population (15). More research is needed on referral practices and potential biases among “gateway providers” (e.g., case work- Limitations ers, probation officers, school counselors, teachers) in the identification of the mental health needs of youths and re- This study addressed many of the limitations of previ- ferral to services (38). Interdisciplinary approaches are ous work in this area by 1) including sufficient representa- needed to support investigation of how organizational, tion of youths from four major racial/ethnic groups; 2) us- policy, and funding factors interact with community and ing well-established instruments to assess service use, family help-seeking preferences to result in differential ac- psychiatric diagnoses, and functional impairment; 3) in- cess and utilization. Newer models of mental health help cluding assessment of most of the other factors known to seeking, such as the network/episode model (33), which predict service use, such as caregiver strain, police con- include greater attention to the social context, may be very tact, and family income; and 4) assessing use of a variety of useful in elucidating the contextual factors that influence types of mental health services. There are, however, some individual family decisions about service utilization. limitations. The service use data relied on parents’ and youths’ self-reports, which may be subject to cognitive Conclusions and culturally influenced biases. Specifically, there may be Racial/ethnic disparities in use of professional mental cultural differences in comfort in disclosing mental health health services are robust. The results of this study rein- service use and/or labeling of service use. In addition, force the public’s and policy makers’ concerns about ra- commonly used school-based services were not included cial/ethnic disparities in mental health service utilization in these analyses. Also, the participation rate was less than for vulnerable youths and families. optimal (67%), but retrospective analyses did not identify major differences in the basic sociodemographic charac- Received April 1, 2003; revisions received Nov. 25, 2003, April 7 teristics of participants, compared to nonparticipants. Al- and May 18, 2004; accepted June 14, 2004. From Child and Adoles- though the sample was diverse, it excluded non-English- cent Services Research Center; the Department of Psychiatry, Univer- sity of California, San Diego, La Jolla, Calif.; the Department of Psy- speaking families and thus may not be representative of all chology, University of California, Los Angeles, Los Angeles; the School families in each of these four major racial/ethnic groups. of Social Work and the Department of Psychology, San Diego State Non-English speakers are likely to be less acculturated to University, San Diego; the Department of Psychology, University of San Diego, San Diego; and the Department of Psychiatry, University the dominant society and less likely to utilize mental of New Mexico, Albuquerque. Address correspondence and reprint health services (32). Thus, exclusion of these families may requests to Dr. Garland, Child and Adolescent Services Research Cen- have actually underestimated the service use disparities ter, Children’s Hospital and Health Center, 3020 Children’s Way (MC 5033), San Diego, CA 92123; [email protected] (e-mail). for some racial/ethnic minority groups. Finally, the sam- The Patterns of Youth Mental Health Care in Public Service Systems ple was a “hybrid,” high-risk sample of youths who had study is supported by NIMH grant MH-55282. Preparation of this ar-

1342 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GARLAND, LAU, YEH, ET AL. ticle was also supported by NIMH grants MH-01544, MH-01767, and 19. Horwitz SM, Hoagwood K, Stiffman AR, Summerfeld T, Weisz JR, MH-01924. Costello EJ, Rost K, Bean DL, Cottler L, Leaf PJ, Roper M, The authors thank Donald Slymen, Ph.D., for statistical consulta- Norquist G: Reliability of the Services Assessment for Children tion and Patricia Wood, M.A., M.P.H., for data analysis assistance. and Adolescents. Psychiatr Serv 2001; 52:1088–1094 20. Shaffer D, Fisher P, Lucas CP, Dulcan MK, Schwab-Stone ME: NIMH Diagnostic Interview Schedule for Children Version IV References (NIMH DISC-IV): description, differences from previous ver- 1. Bui KT, Takeuchi DT: Ethnic minority adolescents and the use sions, and reliability of some common diagnoses. J Am Acad of community mental health care services. Am J Community Child Adolesc Psychiatry 2000; 39:28–38 Psychol 1992; 20:403–417 21. Cantwell DP, Lewinsohn PM, Rohde P, Seely J: Correspondence be- 2. Kataoka SH, Zhang L, Wells KB: Unmet need for mental health tween adolescent report and parent report of psychiatric diagnos- care among US children: variation by ethnicity and insurance tic data. J Am Acad Child Adolesc Psychiatry 1996; 36:610–619 status. Am J Psychiatry 2002; 159:1548–1555 22. Cottler LB, Robins LN, Helzer JE: The reliability of the CIDI-SAM: 3. McCabe K, Yeh M, Hough RL, Landsverk J, Hurlburt MS, Culver a comprehensive substance abuse interview. Br J Addict 1998; SW, Reynolds B: Racial/ethnic representation across five public 84:801–814 sectors of care for youth. J Emotional and Behavioral Disorders 23. Shaffer D, Gould MS, Brasic J, Ambrosini P, Fisher P, Bird H, Alu- 1999; 7:72–82 wahlia S: A Children’s Global Assessment Scale (CGAS). Arch Gen 4. Mental Health: Culture, Race, and Ethnicity—A Supplement to Psychiatry 1983; 40:1228–1231 Mental Health: A Report of the Surgeon General. Rockville, Md, 24. Leaf PJ, Alegria M, Cohen P, Goodman SH, Horwitz SM, Hoven US Department of Health and Human Services, 2001 CW, Narrow WE, Vaden-Kiernan M, Regier DA: Mental health 5. Burns BJ, Costello EJ, Angold A, Tweed D, Stangl D, Farmer EM, service use in the community and schools: results from the Erkanli A: Children’s mental health service use across service four-community MECA study. J Am Acad Child Adolesc Psychia- sectors. Health Aff 1995; 14(3):147–159 try 1996; 35:889–897 6. McKay MM, Pennington J, Lynn C, McCadam K: Understanding 25. Bird HR, Shaffer D, Fisher P, Gould, MS, Stagehezza B, Che JY, urban child mental health service use: two studies of child, Hoven C: The Columbia Impairment Scale (CIS): pilot findings family, and environmental correlates. J Behav Health Serv Res on a measure of global impairment for children and adoles- 2001; 28:475–483 cents. Int J Methods Psychiatr Res 1993; 3:167–176 7. Zahner GEP, Pawelkiewicz W, DeFrancesco JJ, Adnopoz J: Chil- 26. Radloff LS: The CES-D Scale: a self-report depression scale for dren’s mental health service needs and utilization patterns in research in the general population. J Applied Psychol Measure- an urban community: an epidemiological assessment. J Am ment 1977; 1:385–401 Acad Child Adolesc Psychiatry 1992; 31:951–960 27. Stata Reference Manual: Release 7.0. College Station, Tex, Stata 8. Aarons G, Brown S, Hough R, Garland AF, Wood P: Prevalence Corp, 2001 of substance use disorders among adolescents in public sectors 28. Rao JNK, Scott AJ: The analysis of categorical data for complex of care. J Am Acad Child Adolesc Psychiatry 2001; 40:419–426 sample surveys: chi-square tests for goodness of fit and indepen- 9. Garland AF, Hough RL, McCabe KM, Yeh M, Wood P, Aarons GA: dence in two-way tables. J Am Statistical Assoc 1981; 76:221–230 Prevalence of psychiatric disorders in youths across five sectors 29. Kelleher KJ, Moore CD, Childs GE, Angelilli ML, Comer DM: Pa- of care. J Am Acad Child Adolesc Psychiatry 2001; 40:409–418 tient race and ethnicity in primary care management of child 10. Garland AF, Hough R, Landsverk J, Brown S: Multi-sector com- behavior problems: a report from PROS and ASPN. Med Care plexity of systems of care for youth with mental health needs. 1999; 37:1092–1104 Children’s Services: Social Policy, Research, and Practice 2001; 30. Vega WA, Lopez SR: Priority issues in Latino mental health ser- 4:123–140 vices research. Ment Health Serv Res 2001; 3:189–200 11. Hofstra MB, van der Ende J, Verhulst FC: Child and adolescent 31. Costello EJ, Janiszewski S: Who gets treated? factors associated problems predict DSM-IV disorders in adulthood: a 14-year fol- with referral in children with psychiatric disorders. Acta Psychi- low-up of a Dutch epidemiological sample. J Am Acad Child Ad- atr Scand 1990; 81:523–529 olesc Psychiatry 2002; 41:182–189 32. Hough RL, Landsverk JA, Karno M: Utilization of health and 12. Angold A, Messer SC, Stangl D, Farmer EMZ, Costello EJ, Burns BJ: mental health services by Los Angeles Mexican-American and Perceived parental burden and service use for child and adoles- non-Latino whites. Arch Gen Psychiatry 1987; 44:702–709 cent psychiatric disorders. Am J Public Health 1998; 88:75–80 33. Costello EJ, Pescosolido BA, Angold A, Burns BJ: A family net- 13. Brannan AM, Heflinger GA, Bickman L: The Caregiver Strain work-based model of access to child mental health services. Questionnaire: measuring the impact on the family of living with a child with serious emotional disturbance. J Emotional Res Community Ment Health 1998; 9:165–190 and Behavioral Disorders 1997; 5:212–222 34. Sue S, Zane N: The role of culture and cultural techniques in 14. McCabe KM, Yeh M, Lau A, Garland AF, Hough RL: Racial/ethnic psychotherapy: a critique and reformulation. Am Psychol differences in caregiver strain and perceived social support 1987; 42:37–45 among parents of youth with emotional and behavioral prob- 35. Kleinman A: Clinical relevance of anthropological and cross- lems. Ment Health Serv Res 2003; 5:137–147 cultural research: concepts and strategies. Am J Psychiatry 15. Breda C: The mental health orientation of juvenile courts. J Be- 1978; 135:427–431 hav Health Serv Res 2001; 28:89–95 36. Bussing R, Schoenberg NE, Perwien AR: Knowledge and informa- 16. Wordes M, Bynum TS, Corley CJ: Locking up youth: the impact tion about ADHD: evidence of cultural differences among Afri- of race on detention decisions. J Research in Crime and Delin- can-American and white parents. Soc Sci Med 1998; 46:919–928 quency 1994; 31:149–165 37. Yeh M, Hough RL, McCabe K, Lau A, Garland A: Parental beliefs 17. Cohen P, Hesselbart CS: Demographic factors in the use of chil- about the causes of child problems: exploring racial/ethnic dren’s mental health services. Am J Public Health 1993; 83:49–52 patterns. J Am Acad Child Adolesc Psychiatry 2004; 43:605–612 18. Garland AF, Aarons GA, Brown SA, Wood PA, Hough RL: Diag- 38. Stiffman A, Hadley-Ives E, Dore P, Polgar M, Horvath VE, Striley nostic profiles associated with utilization of mental health and C, Elze D: Youths’ access to mental health services: the role of substance abuse services among high-risk youth. Psychiatr Serv providers’ training, resource connectivity, and assessment of 2003; 54:562–564 need. Ment Health Serv Res 2000; 2:141–154

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1343 Article

A 3-Year Panel Study of Mental Disorders Among Adolescents in Taiwan

Susan S.F. Gau, M.D., Ph.D. Objective: This study investigated the stance use disorders in the third. During prevalence and changing trends of mental the 3 years, the rates for ADHD, specific M.Y. Chong, M.D., Ph.D. disorders and the effects of gender and ur- phobia, and social phobia decreased, and banization among adolescents in Taiwan. the rates for major depression and sub- Tony H.H. Chen, M.D., Ph.D. Method: A random sample of seventh- stance use disorders, conversely, in- grade students (N=1,070) was recruited creased. Although conduct disorder, Andrew T.A. Cheng, M.D., Ph.D., from one urban and one rural junior high ADHD, and substance use disorders were school in which 1,051 (98.2%) and 1,035 D.Sc., F.R.C.Psych. more prevalent among boys, the rates for (96.7%) were reassessed in the second and major depression, social phobia, specific third years, respectively. A two-stage case phobia, and adjustment disorder were identification was conducted by mental higher among girls. Rural adolescents had health professionals with the Schedule for higher rates of conduct disorder, opposi- Affective Disorders and Schizophrenia for tional defiant disorder, and substance use School-Age Children—Epidemiologic Ver- disorders than their urban counterparts. sion, supplemented by information from the Child Behavior Checklist. Conclusions: Our findings are similar to Results: The weighted 3-month preva- those of previous studies among adoles- lence rates across the 3 consecutive years cents in prevalence rates, changing trends for overall psychiatric disorders were of most mental disorders, and gender ef- 20.3%, 22.7%, and 14.8%, respectively. fects. The differential changing trends in The most prevalent psychiatric condition various diagnostic groups may imply the was attention deficit hyperactivity disor- importance of specific measures for pre- der (ADHD) in the first 2 years and sub- vention during adolescence.

(Am J Psychiatry 2005; 162:1344–1350)

Previous work has strongly supported the idea that disorder and substance use disorders have differed consid- childhood and adolescent psychopathology are predictive erably across countries and ethnicity (19, 20). of poor adjustment and psychopathology in adulthood (1, Like many other countries, youths in Taiwan today are 2). Knowledge about the magnitude, expression, and facing greater familial-socioenvironmental stress than course of psychiatric illness among children and adoles- their predecessors. Owing to the rigid educational system cents is therefore important. Regarding prevalence, previ- and high parental expectation on academic achievement, ous studies using psychiatric interviews reported that the the competition in joint entrance examinations for senior rates of all disorders varied from 4.6% (3) to 50.4% (4), with high schools (grades 10–12) and universities is very keen an overall rate of around 20% (5, 6). Recent surveys have (21). Following globalization in trade, the restriction of im- generally applied the two-stage design (5, 7, 9) and in- ported duty on alcohol and tobacco has been lessened, cluded the assessment of impairment in the case defini- with increased consumption in recent years (22). More- tion, according to the requirement in DSM-III-R and DSM- over, betel has become more available in Taiwan in the IV (4, 7, 8). past decade because of markedly increased commercial Prevalence rates for attention deficit hyperactivity disor- interest (23). These environmental changes and their im- der (ADHD), specific phobia, social phobia, and separation pact on adolescent psychopathology deserve an adequate anxiety disorder were in the range of 3%–10% (10, 11), inquiry. Therefore, an epidemiological study of mental 0.3%–21.6% (8, 12), 0.2%–9.3% (13, 14), and 0.2%–7.2% (12, disorders in adolescents was conducted in Taiwan. 13), respectively. These childhood-onset disorders were reported to decline in rates over time during adolescence (15, The Taiwan Epidemiological Study 16). Conversely, the rates for conduct disorder/oppositional of Mental Disorders in Adolescents defiant disorder (3%–5%), substance use disorders, major depression (1.3%–7.0%), and dysthymia (0.4%–8.0%) were This project consisted of a cross-sectional and a longitu- found to increase with age during adolescence (8, 17, 18). dinal survey. The cross-sectional survey was conducted The magnitudes of conduct disorder/oppositional defiant among 774 ninth-grade students in 1994–1995, focusing on

1344 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GAU, CHONG, CHEN, ET AL. substance use disorders (9). The two-stage case finding screening items entirely, according to the response from the study strategy was applied with a brief screening tool for any sub- subjects without any personal judgment. Based on the Chinese stance use disorders and the Chinese version of the Sched- K-SADS-E interview, the nine consultant child psychiatrists then made psychiatric diagnoses. The subjects with a positive response ule for Affective Disorders and Schizophrenia for School- to any of the Chinese K-SADS-E screening items were treated as Age Children—Epidemiologic Version (K-SADS-E) (9, 24, screened positive. In this pretest, the overall sensitivity and speci- 25). The lifetime weighted prevalence rate of any DSM-III-R ficity of the screening interview against any Chinese K-SADS-E di- substance use disorder was 11.0% and was significantly agnostic category were calculated to be 78% and 98%, respectively. higher in boys, in a rural community, and in classes with Teacher report form of the Child Behavior Checklist. The poor academic performance (9). A 3-year panel study was Chinese version of the teacher report fform of the Child Behavior then carried out to investigate the prevalence and trend of Checklist has been proved to be a reliable and valid instrument when used in Taiwanese adolescent populations (26). psychiatric disorders in 1995–1997. This article reports the research methods, prevalence rates of specific mental dis- The Fieldwork orders, and their changing trends in relation to gender and The Review Board of the Department of Health, Taiwan, ap- urban or rural settings across the 3 years. proved this study as ethical for studying adolescents. Child assent and oral informed consent were obtained from the study subjects and their parents, respectively, after detailed explanation of the Method purpose and interview procedures of this study, and confidential- ity about interview records was ensured. The fieldwork was then This was a school-based panel study among a sample of seventh- conducted at school following a timetable arranged by tutors of grade students established in the first year of the survey. the study classes. Study Subjects The same threshold used in the pretest for the first-stage screening was employed in the main survey. In the second stage, Eighteen of 34 classes and eight of 10 classes in grade 7 were all of those who were screened positive and every 1 in 10 who was randomly selected from one urban and one rural junior high randomly selected from those who were screened negative were school in South Taiwan, respectively. The two schools were se- immediately given the second-stage Chinese K-SADS-E inter- lected based on the positive response and cooperativeness of view, conducted by child psychiatrists who were blind to the their principals. The former represented a typical urban school screening results. There was no time lag between the first- and with a high pass rate in the joint entrance examination for senior second-stage interviews, and none of the respondents who re- high school; the latter represented a typical rural school with a ceived the screening interview refused to take the diagnostic in- low pass rate in such an examination over the years. All the stu- terview. The same two-stage case-finding procedure was con- dents in the selected classes were included in the study sample. ducted in the next 2 consecutive years. Teacher report forms were The study sample consisted of 1,070 students (532 boys and 538 collected soon after the fieldwork. girls, ages 13–15), in which 725 (67.8%) were from urban areas (357 boys and 368 girls) and 345 (32.2%) were from rural areas Psychiatric Diagnosis (175 boys and 170 girls). The follow-up rates were 98.2% (N= 1,051) and 96.7% (N=1,035) in the second year (eighth grade) and A psychiatric diagnosis of the study subjects was first made by the third year (ninth grade), respectively. No significant differ- child psychiatrists who conducted the Chinese K-SADS-E inter- ences in gender and urban-rural distribution were found be- view according to the DSM-IV. These diagnoses were then inde- tween the respondents and the nonrespondents. pendently reassessed by two senior psychiatrists (S.S.F.G. and A.T.A.C.) by a systematic review of all the interview records. In our Measures reassessment, the principle of rate-down was employed, and any Chinese K-SADS-E. The K-SADS-E is a semistructured clinical information that was dubious or uncertain was discarded. Psychi- interview for the systematic assessment of both past and current atric diagnoses generated from this reassessment were jointly dis- episodes of mental disorders in children and adolescents (24). cussed, and our consensus diagnosis was taken as final. To mini- Development of the Chinese version of the K-SADS-E was carried mize the likely underreporting of externalized disorders by the out by the Child Psychiatry Research Group in Taiwan (25), which study subjects, information regarding behavioral syndromes included a two-stage translation and modification of several gathered from the teacher report forms was incorporated into our items with psycholinguistic equivalents relevant to Taiwanese diagnostic consideration for the best estimation of ADHD, oppo- culture. Further modification to meet the DSM-IV diagnostic cri- sitional defiant disorder, and conduct disorder. teria and an additional section developed for betel use disorder Statistical Analysis were performed by the research team (9). In this panel study, all the screening items in individual sections of the Chinese K-SADS-E Statistical analyses were performed with SAS, version 8.2 (SAS were grouped together to form a separate screening version for Institute, Cary, N.C.). The preselected alpha value was 0.05. The 3- use in the first stage of case finding. month weighted prevalence rates and their variance for individ- The interrater reliability of the Chinese K-SADS-E was exam- ual psychiatric disorders were calculated with the formula ined before administration of the cross-sectional survey among p=λ1π+λ2(1–π) nine consultant child psychiatrists in the research team, and the and results showed generalized kappa coefficients ranging from 0.73 to π π λ λ 2 πλ λ π λ λ 0.96 for all mental disorders included in the Chinese K-SADS-E. V(p)=1/N[ (1– )( 1– 2) + 1(1– 1)+1/ƒ(1– ) 2(1– 2)], Validity of the screening version of the Chinese K-SADS-E was as- where p was the weighted prevalence, π was the proportion of the sessed before the administration of this panel study among 124 sample screened positive, 1–π was the proportion of the sample randomly selected eighth-grade students. The screening interview screened negative, ƒ was the fraction of the sample screened neg- was first carried out by 14 psychiatric clinical staff members ative who were interviewed at the second stage, λ1 was the pro- (nurses, psychologists, social workers, and psychiatric residents), portion of cases among the sample screened positive who were who were instructed to record the presence/absence of all the interviewed, and λ2 was the proportion of cases among the sam-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1345 MENTAL DISORDERS AMONG ADOLESCENTS

TABLE 1. Diagnostic Distribution of DSM-IV Psychiatric Disorders in an Adolescent Sample in Taiwan Seventh Grade (N=1,070) Eighth Grade (N=1,051) Ninth Grade (N=1,035) DSM-IV Diagnoses N%b 95% CIb N%b 95% CIb N%b 95% CIb za Disruptive behavioral disorders 105 9.8 7.2 to 12.4 94 9.1 6.5 to 11.8 59 5.7 4.3 to 7.1 –3.4*** Conduct disorder 27 2.5 1.6 to 3.5 30 2.9 1.9 to 3.9 30 2.9 1.9 to 3.9 0.5 Oppositional defiant disorder 20 1.9 1.1 to 2.7 27 2.8 0.6 to 5.0 13 1.3 0.6 to 1.9 –1.0 Attention deficit hyperactivity disorder 81 7.5 5.1 to 10.0 64 6.1 4.6 to 7.5 34 3.3 2.2 to 4.4 –4.2† Depressive disorders Any depressive disorder 7 0.7 0.2 to 1.1 25 2.6 0.4 to 4.8 50 4.8 2.0 to 7.6 6.0† Major depression 5 0.5 0.1 to 0.9 24 2.5 0.3 to 4.7 46 4.4 1.7 to 7.2 5.9† Dysthymic disorder 2 0.2 –0.1 to 0.4 2 0.2 –0.1 to 0.5 6 0.6 0.1 to 1.0 1.6 Anxiety disorders Any anxiety disorder 99 9.2 6.1 to 12.3 78 7.4 3.7 to 11.1 32 3.1 1.0 to 5.1 –5.7† Generalized anxiety disorder 7 0.7 0.2 to 1.1 3 0.3 0.0 to 0.6 4 0.4 0.0 to 0.8 –0.9 Social phobia 36 3.4 1.3 to 3.0 19 1.8 1.0 to 2.6 21 2.0 0.1 to 4.0 –2.0* Specific phobia 54 5.0 2.8 to 7.3 58 5.6 1.9 to 9.2 7 0.7 0.2 to 1.2 –5.2† Separation anxiety disorder 3 0.3 0.0 to 0.6 1 0.1 –0.1 to 0.3 0 0.0 0.0 to 0.0 –1.8 Panic disorder 2 0.2 –0.1 to 0.4 1 0.1 –0.1 to 0.3 0 0.0 0.0 to 0.0 –1.4 Agoraphobia without panic 2 0.2 –0.1 to 0.4 0 0.0 0.0 to 0.0 0 0.0 0.0 to 0.0 –1.7 Obsessive-compulsive disorder 2 0.2 –0.1 to 0.4 2 0.2 –0.1 to 0.5 2 0.2 –0.1 to 0.5 0.0 Adjustment disorders 14 1.3 –0.7 to 3.3 14 1.3 –0.7 to 3.4 3 0.3 0.0 to 0.6 2.4* Substance use disorders Any substance use disorders 24 2.2 1.4 to 3.1 44 4.2 3.0 to 5.4 55 5.3 4.0 to 6.7 3.7*** Nicotine use disorders Any nicotine use disorders 24 2.2 1.4 to 3.1 43 4.1 2.9 to 5.3 55 5.3 4.0 to 6.7 3.7*** Nicotine-related disorder 18 1.7 0.9 to 2.5 25 2.4 1.5 to 3.3 17 1.6 0.9 to 2.4 –0.1 Nicotine dependence 6 0.6 0.1 to 1.0 18 1.7 0.9 to 2.5 38 3.7 2.5 to 4.8 5.1† Betel use disorders Any betel use disorder 9 0.8 0.3 to 1.4 19 1.8 1.0 to 2.6 28 2.7 1.7 to 3.7 3.2** Betel abuse 8 0.8 0.2 to 1.3 16 1.5 0.8 to 2.3 24 2.3 1.4 to 3.2 2.9** Betel dependence 1 0.1 –0.1 to 0.3 3 0.3 0.0 to 0.6 4 0.4 0.0 to 0.8 1.3 Alcohol use disorders Any alcohol use disorder 5 0.5 0.1 to 0.9 13 1.2 0.6 to 1.9 20 1.9 1.1 to 2.8 3.1** Alcohol abuse 5 0.5 0.1 to 0.9 12 1.1 0.5 to 1.8 19 1.8 1.0 to 2.7 3.0** Alcohol dependence 0 0.0 0.0 to 0.0 1 0.1 –0.1 to 0.3 1 0.1 –0.1 to 0.3 0.9 One substance use disorder 13 1.2 0.6 to 1.9 21 2.0 1.2 to 2.8 21 2.0 1.2 to 2.9 1.4 Two or more substance use disorders 11 1.0 0.4 to 1.6 23 2.2 1.3 to 3.1 33 3.2 2.1 to 4.3 3.4*** Any psychiatric disorder 217 20.3 15.9 to 24.6 239 22.7 16.9 to 28.6 154 14.8 11.1 to 18.5 –3.1** a Cochran-Armitage trend test. b Weighted prevalence. *p<0.05. **p<0.01. ***p<0.001. †p<0.0001. ple screened negative who were interviewed (27). The 95% confi- No such trend was observed for conduct disorder and dence interval for weighted prevalence was also calculated. oppositional defiant disorder. Both depressive disorders The Cochran-Armitage test for trend with z statistics was used (notably major depression) and substance use disorders to examine the difference in rates across the three time points. The exact p value was calculated for the test for trend when the conversely increased across the three grades. Moreover, sample size of any of the cells in the two-by-three contingency ta- the increasing trend was significant, mainly for comorbid ble was less than 5. Chi-square tests or Fisher’s exact test (when substance use disorders and not for any single substance the cell number was less than 5) was used to examine the gender use disorder. and the urban-rural difference. Mantel-Haenszel odds ratios and In grades 7 and 8, the most prevalent diagnostic group the corresponding confidence intervals (CIs) were calculated. When the number of any cell of the two-by-two contingency table was disruptive behavioral disorders (notably ADHD), fol- was less than 5, exact CIs were computed for odds ratios. lowed by anxiety disorders (notably specific phobia), and substance use disorders. In grade 9, although disruptive Results behavioral disorders was still ahead of others, the rate for substance use disorders increased to become the second Prevalence of Psychiatric Disorders leading disorder, and depressive disorders (mainly major Table 1 shows the DSM-IV diagnostic distribution of all depression) increased to become third. Rates for the top panel respondents from seventh to ninth grade with three disorders in grade 9 were, in fact, close to each weighted 3-month prevalence rates and their 95% CIs. other. In all three grades, the most common substance of The overall rates for the seventh and eighth grades were abuse/dependence was nicotine, followed by betel and higher than that for the ninth, with a significant decline alcohol. in trend. Such decline was mainly observed for disruptive behavioral disorders and anxiety disorders, specifically Gender Difference among child-onset disorders (including ADHD, specific Disruptive behavioral disorders (mainly conduct disor- phobia, social phobia, and separation anxiety disorder). der and ADHD) and substance use disorders were more

1346 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GAU, CHONG, CHEN, ET AL.

TABLE 2. Gender Differences in Psychiatric Morbidity Across Three Grades in an Adolescent Sample in Taiwana Seventh-Grade Boys (N=532) Eighth-Grade Boys (N=526) Ninth-Grade Boys (N=517) Versus Girls (N=538) Versus Girls (N=525) Versus Girls (N=518) zb DSM-IV Diagnoses Odds Ratio 95% CI Odds Ratio 95% CI Odds Ratio 95% CI Boys Girls Disruptive behavior disorders 3.4 2.1 to 5.4† 4.3 2.6 to 7.1† 4.2 2.2 to 8.1†–2.7** –2.2* Conduct disorder 4.6 1.7 to 15.6c*** 5.2 1.9 to 17.5c*** 6.8 2.3 to 27.0c† 0.7 –0.3 Oppositional defiant disorder 1.9 0.8 to 4.8 2.7 1.2 to 6.1* 1.6 0.5 to 6.3c –0.9 –0.5 Attention deficit hyperactivity disorder 4.5 2.5 to 7.8† 5.2 2.7 to 10.1*** 4.9 2.0 to 11.9†–3.7*** –2.1* Depressive disorders 0.1 0.0 to 1.2d* 0.2 0.0 to 0.5c*** 0.5 0.3 to 1.0* 4.8† 4.2† Anxiety disorders Any anxiety disorder 1.0 0.6 to 1.4 0.3 0.2 to 0.4† 0.3 0.1 to 0.6** –5.7†–2.3* Generalized anxiety disorder 0.4 0.0 to 2.5c 0.5 0.0 to 9.6c —–1.4 –0.3 Social phobia 1.1 0.6 to 2.2 0.5 0.2 to 1.2 0.3 0.1 to 0.7c** –3.1** 0.5 Specific phobia 0.8 0.5 to 1.4 0.2 0.1 to 0.4† 1.3 0.2 to 9.2c –4.0†–3.5*** Adjustment disorder 0.2 0.0 to 0.8c* 0.2 0.0 to 0.7c** —–1.2 –2.1* Substance use disorders Any substance use disorder 4.0 1.4 to 13.6c** 5.6 2.5 to 12.7† 5.5 2.7 to 11.4† 3.5*** 1.2 Nicotine use disorders 4.0 1.4 to 13.6c** 6.6 2.7 to 15.6† 5.5 2.7 to 11.4† 3.5*** 1.2 Betel use disorders 19.5 1.1 to 336.7d** 40.4 2.4 to 670.6d† 60.4 3.7 to 991.6d† 3.3** — Alcohol use disorders 1.5 0.2 to 18.3c 5.6 1.2 to 52.0c* 5.8 1.7 to 31.2c** 3.2** 0.5 Any psychiatric disorder 1.7 1.2 to 2.3*** 1.1 0.8 to 1.4 1.5 1.1 to 2.1* –2.5* –1.4 a p values are for comparisons between boys and girls by chi-square or Fisher’s exact test. b Cochran-Armitage trend test. c Fisher’s exact test. d Fisher’s exact test, adding 0.5 to each cell to compute odds ratios and 95% CIs. *p<0.05. **p<0.01. ***p<0.001. †p<0.0001. prevalent in boys than in girls, whereas the reverse was ob- Discussion served for depressive disorders across the 3 years (Table 2). Girls had higher rates of anxiety disorders and adjustment Methodology Considerations disorder than boys, particularly, higher rates of specific As one of the few studies (1, 8, 28) that have investigated phobia in grade 8 and social phobia in grade 9. the prevalence and the time trend for adolescent psychiat- The significant changing trends in this panel for most ric morbidities, the present study provided a unique mental disorders over the 3 years were observed in boys opportunity to compare the sex and the urban-rural difference of such morbidities between developing and de- and girls. However, the increasing trend for all substance veloped countries. Our study has employed the two-stage use disorders was only significant in boys. The declined design, and all the clinical interviews were conducted by trend for social phobia was only significant in boys and child mental health professionals with the standardized that for adjustment disorder was only significant in girls. Chinese K-SADS-E with cross-cultural validation and ac- Urban-Rural Difference ceptable interrater reliability. Consensus psychiatric diagnoses were made through independent assessment of the The overall rates of mental disorders were generally Chinese K-SADS-E interview records and a joint discussion higher in rural than in urban youths (Table 3). Compared by two senior research psychiatrists. The response rates at to their urban counterparts, rural adolescents had signifi- phase I and the two follow-ups were very high. cantly higher rates for conduct disorder, oppositional defi- Despite all of these strengths, there are some limitations ant disorder, and substance use disorders over the 3 years. of this study that require careful consideration in the in- In grade 9, specific phobia was more prevalent among ru- terpretation of the findings. First, because of a purposeful ral adolescents; in contrast, social phobia was more com- sampling of study schools, its external validity for other mon among urban adolescents. There was no urban-rural Taiwanese adolescent populations needs to be examined. difference in rates for depressive and other anxiety disor- Second, the psychiatric diagnoses were mainly based on ders in all grades. clinical interviews of study subjects without interviewing their parents. Previous studies have shown low agreement The time trends for most disorders in rural and urban among child, parent, and teacher informants in reporting areas also followed those in the total sample, with some children’s emotional and behavioral problems (29) and the exceptions. The increasing trend for all substance use dis- need to incorporate teachers’ reports into the identifica- orders was significant in rural areas but was less apparent tion of externalizing disorders (29). Therefore, we included in urban areas and only significant for nicotine use disor- the teacher report form to make the best estimates of psy- ders. The declining trend for social phobia was only signif- chiatric diagnoses of ADHD, conduct disorder, and oppo- icant among rural adolescents and that for specific phobia sitional defiant disorder. Finally, in this two-stage case- was greater in urban than in rural areas, notably in grade 9. finding study, despite the fact that all the study subjects

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1347 MENTAL DISORDERS AMONG ADOLESCENTS

TABLE 3. Urban-Rural Difference in Psychiatric Morbidity Across Three Grades in an Adolescent Sample in Taiwana Seventh-Grade Eighth-Grade Ninth-Grade Adolescents: Rural (N=345) Adolescents: Rural (N=342) Adolescents: Rural (N=339) Versus Urban (N=725) Versus Urban (N=709) Versus Urban (N=696) zb DSM-IV Diagnosis Odds Ratio 95% CI Odds Ratio 95% CI Odds Ratio 95% CI Rural Urban Disruptive behavioral disorders 1.5 1.0 to 2.2 1.4 0.9 to 2.2 1.8 1.1 to 3.1* –1.7 –2.9** Conduct disorder 3.2 1.5 to 6.9** 2.8 1.3 to 5.8** 4.3 2.0 to 9.3† 0.7 –0.1 Oppositional defiant disorder 3.2 1.3 to 8.0** 3.3 1.5 to 7.2** 1.8 0.6 to 5.3 –1.2 –0.2 Attention deficit hyperactivity – disorder 1.0 0.6 to 1.7 0.9 0.5 to 1.5 1.1 0.6 to 2.3 –2.3* 3.5*** Depressive disorder 0.8 0.1 to 5.2c 0.4 0.1 to 1.1c 1.1 0.6 to 1.9 3.9† 4.6† Anxiety disorders Any anxiety disorder 1.0 0.6 to 1.5 0.9 0.6 to 1.5 0.6 0.2 to 1.3 –3.8†–4.1† Generalized anxiety disorder —— 0.7 0.0 to 8.6c 1.2 –1.3 Social phobia 1.2 0.6 to 2.4 1.2 0.5 to 3.1 0.2 0.0 to 0.9c* –2.9** –0.6 Specific phobia 1.0 0.6 to 1.8 1.0 0.6 to 1.8 5.2 0.8 to 54.7c* –2.3* –4.7† Substance use disorders Any substance use disorder 15.6 4.6 to 82.0c† 12.2 5.4 to 27.6† 7.4 3.9 to 14.1† 2.8** 2.7** Nicotine use disorders 15.6 4.6 to 82.0c† 11.8 5.2 to 26.8† 7.4 3.9 to 14.1† 2.8** 2.7** Betel use disorders 41.0 2.4 to 705.9d† 39.3 6.1 to 1641.0c† 28.8 7.1 to 251.4c† 3.0** 1.4 Alcohol use disorders 8.5 0.8 to 418.8c* 7.1 1.8 to 40.3c** 12.2 3.5 to 65.3c† 2.9** 1.0 Any psychiatric disorder 1.3 0.9 to 1.7 1.4 1.1 to 1.9* 1.7 1.2 to 2.4** –1.0 –3.0** a p values are for comparison between rural and urban by chi-square or Fisher’s exact test. b Cochran-Armitage trend test. c Fisher’s exact test. d Fisher’s exact test, adding 0.5 to each cell to compute odds ratio and 95% CI. *p<0.05. **p<0.01. ***p<0.001. †p<0.0001. were screened at the first stage for 3 consecutive years, the countries (8, 33), with a nonsignificant increase for con- second-stage psychiatric interview was not performed duct disorder over the 3 years in boys. among all of them. The lack of complete information in A unique finding of this study is that unlike Western so- psychiatric diagnoses for all study subjects has prevented cieties (19), betel instead of alcohol was the second (after us from conducting longitudinal analyses using a multi- nicotine) most prevalent abused substance among our level model to examine the trajectories of psychiatric diag- study subjects, which might be attributable to an in- noses at an individual level. creased availability of betel because of commercial interest in recent years and the popularity of the betel chewing Prevalence and Trend habit, particularly in rural areas (9). The fact that all the The magnitude of total psychiatric morbidities esti- study subjects with substance use disorders had nicotine mated in this study was in the middle of previous studies use disorder and the rates of nicotine dependence had (5, 6) and was similar to those (20.9% and 20.3%) of two markedly increased over time strongly indicates the seri- large-scale epidemiology studies of youths in the United ousness of nicotine damage on health among Taiwanese States (4, 8). However, the clinical interview and diagnostic adolescents today. criteria (DSM-III-R) used in those two U.S. studies were The 3-month prevalence rate of depressive disorders in different from those in this study. Shaffer et al. (4) em- this study was somewhat lower than those in previous ployed the National Institute of Mental Health’s Diagnostic studies, which largely reported 6- or 12-month prevalence Interview Schedule for Children, Version 2.3, and Costello rates in similar age populations (4, 8). However, our find- et al. (7) used the Child and Adolescent Psychiatric Assess- ing of a significant increase in time trend of such morbid- ment. ity in both sexes was consistent with previous work (20). Although there was only a minor change in the diagnos- Consistent with previous studies, specific phobia was tic criteria for ADHD from DSM-III-R to DSM-IV, the aver- the most common anxiety disorder, followed by social age estimated prevalence of ADHD has been reported to phobia, then generalized anxiety disorder (12–14). The increase from 3%–5% with DSM-III-R to 9%–10% with lower trend in the rates of total anxiety disorders mainly DSM-IV with the three newly created subtypes (11, 30). came from specific phobia, social phobia, and separation The overall prevalence of ADHD in this study was close to anxiety disorder (16), and girls had a greater stability of in- the figures in recent studies in Australia (31) with the Diag- ternalized disorders than boys over time (34). nostic Interview Schedule for Children and in Brazil (32) with the 18 DSM-IV ADHD symptom items and clinical di- Gender Difference agnosis. Similar to findings in previous studies (15), the Similar to earlier studies, we found higher rates of ADHD, rates of ADHD declined over adolescence. Our rates of conduct disorder, oppositional defiant disorder, and sub- conduct disorder and oppositional defiant disorder were stance use disorders in boys, contributing to a higher total in the lower part of the reported rates across cultures and psychiatric morbidity than girls. Higher rates of anxiety

1348 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 GAU, CHONG, CHEN, ET AL. disorders in girls, notably specific phobia and social pho- adolescents. Our findings have implied that the ameliora- bia (35), were also observed. The emerging trend of female tion of detrimental risk factors in social environment (be excess in depressive disorders at adolescence reported in it more prevalent in rural or urban settings in different so- previous Western surveys (13, 20, 34, 36) was also evident cieties) for mental disorders in adolescents may serve as in this study. the target for primary prevention. Betel abuse, a specific disorder in Taiwan and certain Asian countries, ought to Urban-Rural Difference be prevented among both adolescent and adult popula- Our findings of higher rates of conduct disorder/oppo- tions, especially in rural areas. Further investigation of sitional defiant disorder and substance use disorders in risk factors, patterns of comorbidity, and the trajectories rural areas, although contrary to that in earlier Western of psychopathology during adolescence is crucial for the studies (18, 33), were in accordance with those from recent identification of the target for primary prevention among studies in Western (37) and Asian societies (9). The finding different vulnerable groups. of a greater increase in time trend for substance use disorders in rural areas in this study awaits further examination Received Oct. 1, 2003; revisions received Feb. 18 and May 11, 2004; in other societies. accepted June 14, 2004. From the Department of Psychiatry, College of Medicine, and the Institute of Preventive Medicine, College of Pub- Urban-rural differences in psychiatric morbidity are lic Health, National Taiwan University, and National Taiwan Univer- likely to be associated with multiple social environmental sity Hospital, Taipei, Taiwan; the Department of Psychiatry, Chang factors. Urban neighborhood as a risk factor for psychiat- Gung Memorial Hospital and Chang Gung University, Kaohsiung, Tai- wan; and the Institute of Biomedical Sciences, Academia Sinica. Ad- ric disorders has been explained by its close association dress correspondence and reprint requests to Dr. Cheng, Institute of with lower socioeconomic status in most cities in devel- Biomedical Sciences, Academia Sinica, Taipei, Taiwan; bmandrew@ oped countries (38). This notion might have at least in part gate.sinica.edu.tw (e-mail). Supported by grants from the Taiwan Department of Health (DOH- explained the higher rates of conduct disorder/opposi- 85-TD-131, DOH-86-TD-104, and DOH-87-TD-1161). tional defiant disorder, substance use disorders, and total The authors thank all the child psychiatric staffs who participated psychiatric morbidities among adolescents in rural Tai- in this study and the Child Psychiatry Research Group in Taiwan, wan, where the socioeconomic status has been generally which provided the Chinese Schedule for Affective Disorders and Schizophrenia for School-Age Children—Epidemiologic Version for lower than in their urban counterparts (9, 39). Another im- use in this study. portant factor for such difference in morbidity in Taiwan may have come from a positive selective migration of the mentally fit from rural to urban cities (39). For example, References rural children from the families of higher socioeconomic 1. Hofstra MB, van der Ende J, Verhulst FC: Child and adolescent status move to urban cities for better educational oppor- problems predict DSM-IV disorders in adulthood: a 14-year fol- tunities, leaving behind those who are more socioeco- low-up of a Dutch epidemiological sample. J Am Acad Child Ad- nomically disadvantaged and less academically compe- olesc Psychiatry 2002; 41:182–189 tent, with a higher vulnerability to both psychological 2. Heijmens Visser JH, van der Ende J, Koot HM, Verhulst FC: Pre- dictors of psychopathology in young adults referred to mental disorders and substance abuse (9). In consequence, per- health services in childhood or adolescence. Br J Psychiatry formance on the Joint Entrance Examination for Senior 2000; 177:59–65 High School is generally better among urban junior high 3. Vikan A: Psychiatric epidemiology in a sample of 1,510 ten- schools than among their rural counterparts (9). year-old children, I: prevalence. J Child Psychol Psychiatry Findings regarding the relationships between low socio- 1985; 26:55–75 economic status and anxiety and mood disorders have 4. Shaffer D, Fisher P, Dulcan MK, Davies M, Piacentini J, Schwab- Stone ME, Lahey BB, Bourdon K, Jensen PS, Bird HR, Canino G, been contradictory (8, 35). The present study did not find Regier DA: The NIMH Diagnostic Interview Schedule for Chil- any significant association between urban or rural resi- dren, version 2.3 (DISC-2.3): description, acceptability, preva- dency, a proxy for socioeconomic status, and the rates of lence rates, and performance in the MECA Study (Methods for ADHD, depressive disorders, and most anxiety disorders. the Epidemiology of Child and Adolescent Mental Disorders Our finding of different time trends for specific phobia Study). J Am Acad Child Adolesc Psychiatry 1996; 35:865–877 and social phobia across urban or rural residency may de- 5. Roberts RE, Attkisson CC, Rosenblatt A: Prevalence of psychopathology among children and adolescents. Am J Psychiatry serve further inquiry. It is likely that environmental expo- 1998; 155:715–725 sure may last longer for specific phobia in rural areas glo- 6. Angold A, Costello EJ: Developmental epidemiology. Epidemiol bally and may be higher regarding the pressure from social Rev 1995; 17:74–82 contact and school performance in urban areas in devel- 7. Costello EJ, Angold A, Burns BJ, Erkanli A, Stangl DK, Tweed DL: oping countries. The Great Smoky Mountains Study of Youth: functional impairment and serious emotional disturbance. Arch Gen Psychiatry Implications for Prevention 1996; 53:1137–1143 8. Costello EJ, Angold A, Burns BJ, Stangl DK, Tweed DL, Erkanli A, It is imperious to provide a protective environment to Worthman CM: The Great Smoky Mountains Study of Youth: prevent childhood-onset emotional and behavioral disor- goals, design, methods, and the prevalence of DSM-III-R disorders and substance-related disorders among vulnerable ders. Arch Gen Psychiatry 1996; 53:1129–1136

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1349 MENTAL DISORDERS AMONG ADOLESCENTS

9. Chong MY, Chan KW, Cheng AT: Substance use disorders among 25. Gau SF, Soong WT: Psychiatric comorbidity of adolescents with adolescents in Taiwan: prevalence, sociodemographic corre- sleep terrors or sleepwalking: a case-control study. Aust NZ J lates and psychiatric co-morbidity. Psychol Med 1999; 29: Psychiatry 1999; 33:734–739 1387–1396 26. Yang HJ, Soong WT, Chiang CN, Chen WJ: Competence and be- 10. Goldman LS, Genel M, Bezman RJ, Slanetz PJ (Council on Scien- havioral/emotional problems among Taiwanese adolescents tific Affairs, American Medical Association): Diagnosis and as reported by parents and teachers. J Am Acad Child Adolesc treatment of attention-deficit/hyperactivity disorder in chil- Psychiatry 2000; 39:232–239 dren and adolescents. JAMA 1998; 279:1100–1107 27. Shrout PE, Skodol AE, Dohrenwend BP: A two-stage approach 11. Wolraich ML, Hannah JN, Pinnock TY, Baumgaertel A, Brown J: for case identification and diagnosis: first-stage instrument, in Comparison of diagnostic criteria for attention-deficit hyperac- Mental Disorders in the Community: Progress and Challenges. tivity disorder in a county-wide sample. J Am Acad Child Ado- Edited by Barret JE, Rose RM. New York, Guilford, 1986, pp lesc Psychiatry 1996; 35:319–324 286–303 12. Simonoff E, Pickles A, Meyer JM, Silberg JL, Maes HH, Loeber R, 28. Newman DL, Moffitt TE, Caspi A, Magdol L, Silva PA, Stanton Rutter M, Hewitt JK, Eaves LJ: The Virginia Twin Study of Adoles- WR: Psychiatric disorder in a birth cohort of young adults: cent Behavioral Development: influences of age, sex, and im- prevalence, comorbidity, clinical significance, and new case in- pairment on rates of disorder. Arch Gen Psychiatry 1997; 54: cidence from ages 11 to 21. J Consult Clin Psychol 1996; 64: 801–808 552–562 13. Lewinsohn PM, Hops H, Roberts RE, Seeley JR, Andrews JA: Ad- 29. Mitsis EM, McKay KE, Schulz KP, Newcorn JH, Halperin JM: Par- olescent psychopathology, I: prevalence and incidence of de- ent-teacher concordance for DSM-IV attention-deficit/hyperac- pression and other DSM-III-R disorders in high school students. tivity disorder in a clinic-referred sample. J Am Acad Child Ado- J Abnorm Psychol 1993; 102:133–144 lesc Psychiatry 2000; 39:308–313 14. Verhulst FC, van der Ende J, Ferdinand RF, Kasius MC: The prev- 30. Baumgaertel A, Wolraich ML, Dietrich M: Comparison of diag- alence of DSM-III-R diagnoses in a national sample of Dutch ad- nostic criteria for attention deficit disorders in a German ele- olescents. Arch Gen Psychiatry 1997; 54:329–336 mentary school sample. J Am Acad Child Adolesc Psychiatry 1995; 34:629–638 15. Biederman J, Faraone S, Milberger S, Curtis S, Chen L, Marrs A, 31. Graetz BW, Sawyer MG, Hazell PL, Arney F, Baghurst P: Validity Ouellette C, Moore P, Spencer T: Predictors of persistence and of DSM-IV ADHD subtypes in a nationally representative sam- remission of ADHD into adolescence: results from a four-year ple of Australian children and adolescents. J Am Acad Child Ad- prospective follow-up study. J Am Acad Child Adolesc Psychia- olesc Psychiatry 2001; 40:1410–1417 try 1996; 35:343–351 32. Rohde LA, Biederman J, Busnello EA, Zimmermann H, Schmitz 16. Last CG, Perrin S, Hersen M, Kazdin AE: A prospective study of M, Martins S, Tramontina S: ADHD in a school sample of Brazil- childhood anxiety disorders. J Am Acad Child Adolesc Psychia- ian adolescents: a study of prevalence, comorbid conditions, try 1996; 35:1502–1510 and impairments. J Am Acad Child Adolesc Psychiatry 1999; 17. Merikangas KR, Avenevoli S: Epidemiology of mood and anxi- 38:716–722 ety disorders in children and adolescents, in Textbook in Psy- 33. Offord DR, Boyle MH, Szatmari P, Rae-Grant NI, Links PS, Cad- chiatric Epidemiology. Edited by Tsuang M, Tohen M. New man DT, Byles JA, Crawford JW, Blum HM, Byrne C: Ontario York, Wiley-Liss, 2002, pp 657–704 Child Health Study, II: six-month prevalence of disorder and 18. Rutter M, Cox A, Tupling C, Berger M, Yule W: Attainment and rates of service utilization. Arch Gen Psychiatry 1987; 44:832– adjustment in two geographical areas, I: the prevalence of psy- 836 chiatric disorder. Br J Psychiatry 1975; 126:493–509 34. McGee R, Feehan M, Williams S, Anderson J: DSM-III disorders 19. Weinberg NZ, Rahdert E, Colliver JD, Glantz MD: Adolescent from age 11 to age 15 years. J Am Acad Child Adolesc Psychia- substance abuse: a review of the past 10 years. J Am Acad Child try 1992; 31:50–59 Adolesc Psychiatry 1998; 37:252–261 35. Velez CN, Johnson J, Cohen P: A longitudinal analysis of se- 20. Hankin BL, Abramson LY, Moffitt TE, Silva PA, McGee R, Angell lected risk factors for childhood psychopathology. J Am Acad KE: Development of depression from preadolescence to young Child Adolesc Psychiatry 1989; 28:861–864 adulthood: emerging gender differences in a 10-year longitu- 36. Kessler RC, Walters EE: Epidemiology of DSM-III-R major de- dinal study. J Abnorm Psychol 1998; 107:128–140 pression and minor depression among adolescents and young 21. Gau SF, Soong WT: Sleep problems of junior high school stu- adults in the National Comorbidity Survey. Depress Anxiety dents in Taipei. Sleep 1995; 18:667–673 1998; 7:3–14 22. Taiwan Tobacco and Wine Statistical Yearbook. Taipei, Taiwan 37. Atav S, Spencer GA: Health risk behaviors among adolescents Tobacco and Wine Board, 1986 and 2001 attending rural, suburban, and urban schools: a comparative 23. Ho CS, Gee MJ, Tsai CC, Lo CI, Hwang MN: Factors related to be- study. Fam Community Health 2002; 25:53–64 tel chewing among junior high school students in Taiwan. 38. Paykel ES, Abbott R, Jenkins R, Brugha TS, Meltzer H: Urban-ru- Community Dent Oral Epidemiol 2000; 28:150–154 ral mental health differences in Great Britain: findings from 24. Puig-Antich J, Chambers WJ: The Schedule for Affective Disor- the National Morbidity Survey. Psychol Med 2000; 30:269–280 ders and Schizophrenia for School-Age Children—Epidemio- 39. Cheng TA: Urbanisation and minor psychiatric morbidity: a logic Version (K-SADS-E). New York, New York State Psychiatric community study in Taiwan. Soc Psychiatry Psychiatr Epide- Institute, Biometrics Research, 1978 miol 1989; 24:309–316

1350 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

A Randomized, Double-Blind, Placebo-Controlled Trial of Quetiapine in the Treatment of Bipolar I or II Depression

Joseph R. Calabrese, M.D. Objective: There is a major unmet need portions of patients meeting response cri- for effective options in the treatment of teria (≥50% Montgomery-Åsberg Depres- Paul E. Keck, Jr., M.D. bipolar depression. sion Rating Scale score improvement) at Method: Five hundred forty-two outpa- the final assessment in the groups taking 600 and 300 mg/day of quetiapine were Wayne Macfadden, M.D. tients with bipolar I (N=360) or II (N=182) disorder experiencing a major depressive 58.2% and 57.6%, respectively, versus episode (DSM-IV) were randomly assigned 36.1% for placebo. The proportions of pa- Margaret Minkwitz, Ph.D. to 8 weeks of quetiapine (600 or 300 mg/ tients meeting remission criteria (Mont- day) or placebo. The primary efficacy gomery-Åsberg Depression Rating Scale Terence A. Ketter, M.D. measure was mean change from baseline ≤12) were 52.9% in the groups taking 600 to week 8 in the Montgomery-Åsberg De- and 300 mg/day of quetiapine versus Richard H. Weisler, M.D. pression Rating Scale total score. Addi- 28.4% for placebo. Quetiapine at 600 and tional efficacy assessments included the 300 mg/day significantly improved 9 of 10 Hamilton Depression Rating Scale, Clini- and 8 of 10 Montgomery-Åsberg Depres- Andrew J. Cutler, M.D. cal Global Impression of severity and im- sion Rating Scale items, respectively, com- provement, Hamilton Anxiety Rating pared to placebo, including the core Scale, Pittsburgh Sleep Quality Index, and Robin McCoy, R.N. symptoms of depression. Treatment- Quality of Life Enjoyment and Satisfaction emergent mania rates were low and simi- Questionnaire. Ellis Wilson, M.S. lar for the quetiapine and placebo groups Results: Quetiapine at either dose dem- (3.2% and 3.9%, respectively). Jamie Mullen, M.D. onstrated statistically significant improvement in Montgomery-Åsberg Depression Conclusions: Quetiapine monotherapy Rating Scale total scores compared with is efficacious and well tolerated for the The BOLDER Study Group placebo from week 1 onward. The pro- treatment of bipolar depression.

(Am J Psychiatry 2005; 162:1351–1360)

Depressive episodes in bipolar I and II disorder are an More recently, the atypical antipsychotic olanzapine important source of morbidity and mortality. While symp- was found to be superior to placebo in the treatment of tomatic, patients with bipolar I disorder experience de- acute bipolar I depression as monotherapy when data pressive symptoms for about threefold longer than manic were pooled from two 8-week trials (14). Fixed doses of symptoms, and the recovery time is considerably longer olanzapine in combination with the antidepressant flu- for depressive than manic episodes (1–4). Symptomatic oxetine were administered to small groups of patients in patients with bipolar II disorder spend almost 40 times these studies and were found to be both superior to pla- longer depressed than hypomanic patients (5). Bipolar de- cebo and superior to olanzapine monotherapy. pression is associated with high rates of disability (6) and Quetiapine is efficacious in the treatment of acute bipo- an increased risk of suicide, which occurs in 10% to 20% of lar mania, both as monotherapy and in combination with other mood stabilizers (15, 16). Preliminary evidence for patients with bipolar disorder (7). the efficacy of quetiapine in the treatment of depressive Although multiple agents, including several atypical an- symptoms in a variety of psychotic and mood disorders tipsychotics, have demonstrated efficacy in the treatment (including bipolar disorder, rapid-cycling bipolar disor- of the manic phase of bipolar I disorder (8), the acute der, and adolescent mania) has been reported in several treatment of bipolar depression has not been as well stud- randomized or open-label studies (17–24). ied (9). Lithium and lamotrigine are recommended as ini- Based on the need for new treatment options for bipolar tial treatments for acute bipolar I depression (10, 11). depression, the effectiveness of atypical antipsychotics in However, the response of bipolar depression to lithium is acute mania, and the emerging evidence for their use in often incomplete in a substantial proportion of patients bipolar depression, we evaluated the efficacy and safety of (12), and the efficacy of lamotrigine in the treatment of quetiapine compared with placebo in the treatment of de- acute bipolar I depression has only been demonstrated in pressive episodes in patients with bipolar I or bipolar II one adequately powered placebo-controlled trial (13). disorder.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1351 QUETIAPINE FOR BIPOLAR DEPRESSION

Method sessments were conducted. The use of all other psychotropic drugs was prohibited during the study. This double-blind, randomized, fixed-dose, placebo-controlled, parallel-group monotherapy study of quetiapine versus Efficacy Evaluations placebo was conducted at 39 centers in the United States between Clinical assessments were conducted at baseline and weekly September 2002 and October 2003. After a washout period of at from weeks 1 to 8. The primary efficacy variable was the mean least five half-lives of any prior psychotropic medications, sub- change in the Montgomery-Åsberg Depression Rating Scale total jects were treated for 8 weeks to evaluate the efficacy, safety, and score from baseline to week 8 (27). tolerability of 600 and 300 mg/day of quetiapine and placebo in Additional efficacy evaluations included a change from base- the treatment of depressive episodes in adult patients with bipo- line to each assessment on the Montgomery-Åsberg Depression lar I or II disorder. Rating Scale, the proportion of patients who achieved a protocol- The study was approved by institutional review boards for each defined response (≥50% reduction from baseline score on the site and performed in accordance with the current amendment of Montgomery-Åsberg Depression Rating Scale), the time to re- the Declaration of Helsinki and the International Conference on sponse, the proportion of patients who achieved remission Harmonization/Good Clinical Practice guidelines. Written in- (Montgomery-Åsberg Depression Rating Scale score ≤12), the formed consent was obtained from all subjects before participa- time to remission, as well as a Montgomery-Åsberg Depression tion. Rating Scale item analysis. The change from baseline to each assessment on the Hamilton depression scale, the Clinical Global Patient Population Impression (CGI) (28) severity of illness score, and the CGI im- Outpatients ages 18 to 65 years who met DSM-IV criteria for bi- provement score were also assessed. polar I or II disorder and were experiencing a major depressive The effect of quetiapine on anxiety symptoms was assessed episode were eligible for inclusion in the study. The diagnosis was with the Hamilton Anxiety Rating Scale (29). Mean change from confirmed with the Structured Clinical Interview for DSM-IV. The baseline to each assessment and at week 8 in the Hamilton anxi- patients were required to have a Hamilton Depression Rating ety scale total score was determined. Scale 17-item score ≥20 (25), a Hamilton depression scale item 1 Quality of sleep was assessed with the Pittsburgh Sleep Quality score ≥2, and a Young Mania Rating Scale (26) score ≤12 at both Index, which measures several dimensions of sleep, including the screening and randomization visits. Inclusion criteria were quality, latency, duration, efficiency, use of medication, and day- based on the Hamilton depression scale rather than the primary time dysfunction (30). efficacy measure (the Montgomery-Åsberg Depression Rating The 16-item short form of the Quality of Life Enjoyment and Scale [27]). Satisfaction Questionnaire was used to measure satisfaction with Patients were excluded from the study if they were diagnosed various areas of daily functioning, such as social relationships, with an axis I disorder other than bipolar disorder that was the living/housing, physical health, medication, and global satisfac- primary focus of treatment within 6 months before the screening, tion (31). The Pittsburgh Sleep Quality Index and the Quality of if the current episode of depression exceeded 12 months or was Life Enjoyment and Satisfaction Questionnaire were adminis- less than 4 weeks in duration, or if they had a history of nonre- tered at baseline and at weeks 4 and 8. sponse to an adequate (6-week) trial of more than two classes of antidepressants during the current episode. Additional exclusion Safety and Tolerability Evaluations criteria included a diagnosis of substance dependence (DSM-IV) Safety and tolerability were evaluated by assessing the inci- or substance use (except for nicotine) within 12 months before dence and severity of adverse events, as well as withdrawals be- the screening or a clinically significant medical illness. Patients cause of adverse events. Extrapyramidal symptoms were assessed who posed a current serious suicidal or homicidal risk were also with the Simpson-Angus Rating Scale (32), and akathisia was as- excluded. Patients were not permitted to take benzodiazepines sessed with the Barnes Rating Scale for Drug-Induced Akathisia during the washout period, and only limited use was permitted (33) at random assignment and at week 8. Measurements of vital during the first 3 weeks after random assignment. signs, including weight and fasting serum glucose levels, were ob- Random assignment was achieved in a non-center-specific tained at each study visit. Twelve-lead ECGs, clinical chemistry, manner with an interactive voice-response central randomiza- and hematology assessments were performed at the screening tion service. Random assignment was stratified according to bi- and at week 8. polar type (I or II) to ensure a relative balance in the total number The incidence of treatment-emergent mania was evaluated by of patients among groups (1:1:1). The patients were randomly as- comparing the percentage of patients in each group who had a to- signed to one of three groups: quetiapine, 600 mg/day; quetia- tal Young Mania Rating Scale score of ≥16 on any two consecutive pine, 300 mg/day; or placebo. visits or at the final assessment, or an adverse event of mania or hypomania. Study Medication Quetiapine (600 mg/day or 300 mg/day) or placebo was ad- Statistical Analyses ministered orally, in a single dose, once a day at bedtime. Que- Primary and secondary efficacy analyses were performed on tiapine was initiated at 50 mg/day and administered to achieve a the intent-to-treat population, which included all randomly as- target dose of 300 mg/day by day 4 or 600 mg/day by week 1. All signed patients who took at least one dose of study medication packaging of treatments was identical, with placebo and active and had at least one postbaseline efficacy assessment. A last-ob- tablets identical in appearance and number. servation-carried-forward analysis was used to impute missing data for patients who withdrew during the study. All statistical Prior and Concomitant Medication tests were two-tailed. The primary analysis of change from base- Nonpsychotropic medication, including over-the-counter line to final assessment in the Montgomery-Åsberg Depression medications taken before entry into the study could be contin- Rating Scale total scores tested the superiority of each dose of ued. Zolpidem tartrate (5–10 mg/day at bedtime for insomnia) quetiapine in the intent-to-treat group (patients with bipolar I or and lorazepam (1–3 mg/day for severe anxiety) were permitted at bipolar II disorder) with an analysis of covariance (ANCOVA) with the discretion of the investigator and only during the first 3 weeks the baseline Montgomery-Åsberg Depression Rating Scale as the of treatment but were withheld for 8 hours before psychiatric as- covariate and included treatment and diagnosis strata as fixed ef-

1352 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 CALABRESE, KECK, MACFADDEN, ET AL.

TABLE 1. Baseline Demographic Characteristics of FIGURE 1. Disposition of Outpatients with Bipolar I or II Screened Outpatients With Bipolar I or II Disorder Who Ex- Disorder Who Experienced a Major Depressive Episode perienced a Major Depressive Episode Patients Who Were Patients screened (N=838) Patients Who Did Not Randomly Assigned Pass Screening to Treatment (N=296) (N=539)a Characteristic N%N% Patients selected for random assignment (N=542) Bipolar I patients (N=360) Female sex 168 56.8 308 57.1 Bipolar II patients (N=182) Caucasian race 227 76.7 438 81.3 Age (years) 18–39 163 55.1 318 59.0 40–59 122 41.2 310 39.0 Patients randomly Patients randomly Patients randomly ≥60 10 3.4 5 0.9 assigned to assigned to assigned to a quetiapine, quetiapine, placebo Safety population that excluded three patients who did not receive 600 mg/day 300 mg/day (N=181) any dose of study medication. (N=180) (N=181) fects in the model, with adjustment for multiple comparisons. Ef- Discontinued Discontinued Discontinued fect size (improvement of quetiapine over placebo divided by (N=82) (N=60) (N=74) pooled standard deviation) was determined with a mixed-model Lost to follow-up Lost to follow-up Lost to follow-up repeated-measures analysis. (N=21) (N=12) (N=11) Differences in response rates between treatment and placebo Adverse event Adverse event Adverse event groups and in patients with and without rapid cycling were as- (N=47) (N=29) (N=16) sessed with a Cochran-Mantel-Haenszel chi-square test across Protocol Protocol Protocol diagnostic strata. Hamilton depression scale, CGI severity and noncompliance noncompliance noncompliance improvement, Young Mania Rating Scale, Hamilton anxiety scale, (N=4) (N=10) (N=11) Pittsburgh Sleep Quality Index, and Quality of Life Enjoyment and Withdrew Withdrew Withdrew Satisfaction Questionnaire scores were tested with ANCOVAs. All informed informed informed consent (N=6) consent (N=5) consent (N=12) secondary analyses were conducted at the nominal significance Lack of efficacy Lack of efficacy Lack of efficacy level of 0.05, with no adjustment for multiple comparisons. (N=1) (N=4) (N=24) Sample sizes were determined to provide 85% power to detect Other (N=3) a difference of 3.6 points on the Montgomery-Åsberg Depression Rating Scale with two-tailed pairwise comparisons between treatment groups and placebo at an alpha level of 0.025 in the intent- Completed study Completed study Completed study to-treat population (patients with bipolar I or bipolar II disorder). (N=98) (N=121) (N=107) Exploratory analyses were carried out on the bipolar I and II subgroups whose group size was not predetermined to provide power for significance testing. Exploratory analyses were limited one dose of study medication and were included in the to descriptions of the mean changes in primary outcome measure across the three treatment groups, and effect size determina- safety population. Of these, 511 had at least one postbase- tions for the groups taking 600 and 300 mg/day of quetiapine. The line assessment and were analyzed for efficacy in the in- repeated measures mixed-effects model included terms for treat- tent-to-treat population. ment, bipolar diagnosis, treatment-by-bipolar diagnosis, base- There were no statistically significant differences be- line Montgomery-Åsberg Depression Rating Scale total score, visit (week), and treatment-by-visit effects. Several covariance tween treatment groups with respect to any demographic structures were examined, including autoregressive, banded and baseline disease characteristic (Table 2). The mean Toeplitz, compound symmetry, and unstructured. The best-fit- age was approximately 37 years, and 58.2% of the patients ting covariance structure, the banded Toeplitz, was determined were women. Mean Montgomery-Åsberg Depression Rat- with the Bayesian information criterion. ing Scale scores at baseline were consistent with moderate Results to severe depression (34). There were no statistically significant differences be- Patients and Disposition tween the quetiapine groups and placebo in the propor- A total of 838 patients were screened, and 542 patients tion of the patients who completed the study: 54% in the with bipolar I (N=360) or bipolar II (N=182) disorder were 600 mg/day quetiapine group, 67% in the 300 mg/day randomly assigned to receive quetiapine, 600 mg/day (N= quetiapine group, and 59% in the placebo group. The most 180); quetiapine, 300 mg/day (N=181); or placebo (N= common reasons for withdrawal were related to adverse 181). There were no significant differences between the events in the quetiapine groups (26.1% and 16.0%) and baseline characteristics of patients who did not pass the lack of efficacy in the placebo group (13.3%). screening compared with those who were randomly as- The use of lorazepam and zolpidem (permitted during signed (Table 1). The most common reason for the screen- the first 3 weeks of the study) was generally low across ing failure was failure to meet eligibility criteria. Figure 1 groups. Lorazepam use during the study was 5.6% and illustrates the disposition of patients during the study. Of 9.5% in the 600 and 300 mg/day quetiapine groups, re- the 542 randomly assigned patients, 539 received at least spectively, compared with 8.3% in the placebo group.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1353 QUETIAPINE FOR BIPOLAR DEPRESSION

TABLE 2. Baseline Demographic and Clinical Characteristics of Outpatients With Bipolar I or II Disorder Who Experienced a Major Depressive Episodea

Patients Taking Quetiapine Patients Taking Placebo Characteristic 600 mg/day (N=170) 300 mg/day (N=172) (N=169) N % N % N % Sex Male 71 41.8 79 45.9 64 37.9 Female 99 58.2 93 54.1 105 62.1 Race Caucasian 144 84.7 141 82.0 129 76.3 Black 1810.62313.42615.4 Hispanic 5 2.9 7 4.1 9 5.3 Other 3 1.8 1 0.6 5 2.9 DSM-IV diagnosis Bipolar I disorder 114 67.1 116 67.4 112 66.3 Bipolar II disorder 56 32.9 56 32.6 57 33.7 DSM-IV rapid cycling 3118.24224.43520.7

Mean SD Mean SD Mean SD

Age (years) 37.3 11.4 36.6 11.2 38.3 11.1 Baseline scores Montgomery-Åsberg Depression Rating Scale 30.3 5.3 30.4 5.0 30.6 5.3 Hamilton Depression Rating Scale 24.7 3.5 24.5 3.0 24.6 3.3 Hamilton Anxiety Rating Scale 18.7 7.3 18.6 7.3 18.9 7.3 a Intent-to-treat analysis.

FIGURE 2. Least-Squares Mean Change From Baseline in scores were 30.3 (SD=5.3), 30.4 (SD=5.0), and 30.6 (SD=5.3) Montgomery-Åsberg Depression Rating Scale Total Score at in the 600 mg/day, 300 mg/day, and placebo groups, re- Each Assessment of Outpatients With Bipolar I or II Disor- der Who Experienced a Major Depressive Episodea spectively. Quetiapine at a dose of either 600 or 300 mg/ day demonstrated significantly greater mean improve- Patients assigned to placebo (N=169) ment in Montgomery-Åsberg Depression Rating Scale to- Patients assigned to quetiapine, 300 mg/day (N=172) Patients assigned to quetiapine, 600 mg/day (N=170) tal scores compared with placebo as early as week 1 and at all time points that followed in the intent-to-treat group of 0 patients with bipolar I or II depression (p<0.001 for both quetiapine doses versus placebo) (Figure 2). The mean change in Montgomery-Åsberg Depression Rating Scale –5 total score from baseline to last assessment was –16.73 in the 600 mg/day group and –16.39 in the 300 mg/day a group, compared with –10.26 in the placebo group –10 a < a (p 0.001 for both quetiapine doses versus placebo) (Table 3, Figure 2). The effect sizes were 0.81 for 600 mg/day and a a 0.67 for 300 mg/day of quetiapine. a a –15 a a a a Approximately 58% of the patients treated with either a a a a dose of quetiapine were responders at the final assess-

Mean Change From Baseline in From Mean Change a ment, and both doses resulted in significantly higher re- –20 sponse rates than placebo (36.1%) (p<0.001). Notably, the 0 12345678 percentage of patients meeting response criteria with 600 Montgomery-Åsberg Depression Scale Total Score Scale Total Montgomery-Åsberg Depression Study Week mg/day of quetiapine was significantly higher as early as < a Intent-to-treat, last-observation-carried-forward analyses. Improve- week 1 (24.3%) versus placebo (10.7%) (p 0.001). In the ment in Montgomery-Åsberg Depression Rating Scale total score group taking 300 mg/day of quetiapine, a significantly with both doses of quetiapine (600 mg/day and 300 mg/day) was higher response rate (37.2%) versus placebo (19.5%) was significantly greater than placebo at every assessment (p<0.001). apparent by week 2 (p<0.001). The median time to response was significantly shorter for both 600 mg/day (22 Zolpidem use during the study was 6.7% and 4.5% in the days) and 300 mg/day (22 days) of quetiapine compared 600 and 300 mg/day quetiapine groups, respectively, com- with placebo (36 days) (log-rank χ2=33.1, df=2, p<0.001). pared with 8.3% in the placebo group. The percentage of patients meeting remission criteria at the final assessment was 52.9% in both the groups taking Efficacy 600 and 300 mg/day of quetiapine, significantly higher Montgomery-Åsberg Depression Rating Scale. Mean than the placebo rate of 28.4% in each group (p<0.001). baseline Montgomery-Åsberg Depression Rating Scale The median time to remission was significantly shorter for

1354 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 CALABRESE, KECK, MACFADDEN, ET AL.

TABLE 3. Baseline and Mean Change in Efficacy Measures at the Last Assessment of Outpatients With Bipolar I or II Disorder Who Experienced a Major Depressive Episodea

Baseline Score Change in Score Analysis (comparison with placebo) Measure and Treatment Mean SDat Last Assessment ANCOVA (df=1)b p Montgomery-Åsberg Depression Rating Scale 600 mg/day of quetiapine 30.3 5.3 –16.73 −6.47 (1.12) <0.001 300 mg/day of quetiapine 30.4 5.0 –16.39 −6.13 (1.12) <0.001 Placebo 30.6 5.3 –10.26 Hamilton Depression Scale 600 mg/day of quetiapine 24.7 3.5 –13.84 −5.29 (0.81) <0.001 300 mg/day of quetiapine 24.5 3.0 –13.38 −4.84 (0.80) <0.001 Placebo 24.6 3.3 –8.54 Hamilton Depression Scale item 1 600 mg/day of quetiapine 2.9 0.5 –1.68 −0.57 (0.12) <0.001 300 mg/day of quetiapine 2.9 0.5 –1.65 −0.54 (0.12) <0.001 Placebo 2.9 0.4 –1.11 Clinical Global Impression scale Improvement 600 mg/day of quetiapine 4.5 0.6 2.37 −0.60 (0.14) <0.001 300 mg/day of quetiapine 4.4 0.5 2.27 −0.71 (0.14) <0.001 Placebo 4.4 0.6 2.97 Severity 600 mg/day of quetiapine 4.5 0.6 –1.66 −0.72 (0.14) <0.001 300 mg/day of quetiapine 4.4 0.5 –1.63 −0.68 (0.14) <0.001 Placebo 4.4 0.6 –0.95 Hamilton Anxiety Rating Scale 600 mg/day of quetiapine 18.7 7.3 –8.75 −3.20 (0.76) <0.001 300 mg/day of quetiapine 18.6 7.3 –8.64 −3.10 (0.76) <0.001 Placebo 18.9 7.3 –5.54 Pittsburgh Sleep Quality Index 600 mg/day of quetiapine 11.6 4.2 –5.46 −2.52 (0.43) <0.001 300 mg/day of quetiapine 11.4 3.8 –5.16 −2.22 (0.44) <0.001 Placebo 11.7 3.8 –2.94 Quality of Life Enjoyment and Satisfaction Questionnaire 600 mg/day of quetiapine 34.1 8.2 11.71 5.27 (1.14) <0.001 300 mg/day of quetiapine 36.1 7.9 10.77 4.33 (1.15) <0.001 Placebo 34.2 7.4 6.44 a Intent-to-treat, last-observation-carried-forward analyses. b Test, treatment contrast within the framework of the ANCOVA, estimated difference (standard error). both 600 mg/day (27 days) and 300 mg/day (29 days) of group taking 600 mg/day group of quetiapine and –16.91 quetiapine compared with placebo (65 days) (log-rank χ2= in the 300 mg/day group, compared with –9.24 in the pla- 32.8, df=2, p<0.001). cebo group (p<0.001 for both quetiapine doses versus pla- Nine out of 10 Montgomery-Åsberg Depression Rating cebo). The effect size in the bipolar I subgroup was 1.09 for Scale items were significantly improved from baseline those assigned to 600 mg/day and 0.91 for those taking 300 compared with placebo in the 600 mg/day quetiapine mg/day of quetiapine. In the subgroup of patients with bi- group, as were eight items in the 300 mg/day quetiapine polar II disorder, the mean change in Montgomery-Åsberg group (p<0.05) (Figure 3). With both doses of quetiapine, Depression Rating Scale total score from baseline to last these items included the core mood symptoms of appar- assessment was smaller than in bipolar I patients. Al- ent sadness, reported sadness, inability to feel, pessimistic though the change in Montgomery-Åsberg Depression thoughts, and suicidal thoughts. The core mood symp- Rating Scale total score from baseline in the patients with toms of apparent sadness, reported sadness, and pessi- bipolar II disorder was statistically superior to placebo at mistic thoughts were significantly improved in both que- most assessments, it did not reach statistical significance tiapine groups as early as week 1 compared with placebo at the final assessment: –14.06 in the group taking 600 mg/ (p<0.05). An inability to feel and suicidal thoughts were day of quetiapine and –14.78 in the group taking 300 mg/ also significantly improved by week 1 in the group taking day compared with –12.35 in the placebo group. The effect 600 mg/day of quetiapine compared with placebo size in the bipolar II subgroup was 0.39 in the 600 mg/day (p<0.05). Both doses of quetiapine were more effective group and 0.28 in the 300 mg/day group. than placebo in reducing suicidal thoughts at the final as- Significant improvement in Montgomery-Åsberg De- sessment (p≤0.001); the reductions with quetiapine were pression Rating Scale total scores compared with placebo approximately twice that of placebo. at the final assessment occurred with quetiapine treat- In the bipolar I subgroup of patients, the mean change ment regardless of the presence of rapid cycling in the in- in Montgomery-Åsberg Depression Rating Scale total tent-to-treat group (patients with bipolar I or II disorder). score from baseline to last assessment was –18.05 in the The mean change in Montgomery-Åsberg Depression Rat-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1355 QUETIAPINE FOR BIPOLAR DEPRESSION

FIGURE 3. Mean Percent Change From Baseline in Individ- –18.6 in the 300 mg/day quetiapine group versus –9.9 in ual Montgomery-Åsberg Depression Rating Scale Items the placebo group (p<0.01 for both quetiapine doses ver- for Outpatients with Bipolar I or II Disorder Experiencing a Major Depressive Episodea sus placebo). The mean change in Montgomery-Åsberg Depression Rating Scale total score at week 8 in the pa- Patients assigned to quetiapine, 600 mg/day (N=170) tients without rapid cycling was –16.6 in the 600 mg/day Patients assigned to quetiapine, 300 mg/day (N=172) group and –15.7 in the 300 mg/day group versus –10.3 in Patients assigned to placebo (N=169) the placebo group (p<0.001 for both quetiapine doses versus placebo). A more detailed analysis of patients with and b Apparent without rapid cycling in this study will be described in a b sadness separate report. In order to explore the role of somnolence or sedation b Reported on efficacy, the mean change from baseline in Montgom- b sadness ery-Åsberg Depression Rating Scale total scores in the patients with and without these adverse events were com- d Inner pared. The number of patients in the intent-to-treat group c tension with reported somnolence/sedation was 195 (57%) for the quetiapine groups combined and 24 (14%) for the placebo b Reduced group. The mean change in the Montgomery-Åsberg De- sleep b pression Rating Scale total score at week 8 in the patients with somnolence/sedation (either bipolar I or II disorder) was –18.8 in the pooled quetiapine groups (600 or 300 mg/ Reduced appetite day) versus –18.9 in the placebo group. In the patients without somnolence/sedation, the mean change in the c Montgomery-Åsberg Depression Rating Scale total score Concentration b difficulties was –19.3 and –11.7 for in the pooled quetiapine and placebo groups, respectively. The placebo group response

d was higher in the patients reporting somnolence/seda- Lassitude tion, but the results with quetiapine were similar in the patients with or without somnolence/sedation.

c Inability Hamilton depression scale. Mean baseline Hamilton b to feel depression scale scores were 24.7 (SD=3.5), 24.5 (SD=3.0), and 24.6 (SD=3.3) in the 600 mg/day, 300 mg/day, and

Pessimistic b placebo groups, respectively (Table 2). Quetiapine at a thoughts b dose of either 600 or 300 mg/day demonstrated significantly greater mean improvements in Hamilton depression scale total scores compared to placebo as early as Suicidal c b thoughts week 1 and at all time points that followed in the patients with bipolar I or II depression (p<0.001). The mean change 0 1020304050 60 70 80 from baseline in Hamilton depression scale scores at Mean Percent Change in Score on week 8 was –13.84, –13.38, and –8.54 in the 600 mg/day, 300 Montgomery-Åsberg Depression Scale Item mg/day, and placebo groups, respectively (p<0.001 for a Intent-to-treat, last-observation-carried-forward analyses. Nine of both quetiapine doses versus placebo). At the end of the 10 and 8 of 10 Montgomery-Åsberg Depression Rating Scale items (including the core mood symptoms of depression [item 1: appar- study, the effect sizes for the group of patients with bipolar ent sadness; item 2: reported sadness; item 8: inability to feel; item I or II disorder with the Hamilton depression scale was 0.93 9: pessimistic thoughts; item 10: suicidal thoughts]) were signifi- for 600 mg/day and 0.74 for 300 mg/day of quetiapine. cantly improved from baseline compared to placebo in the groups taking 600 mg/day and 300 mg/day of quetiapine, respectively Significant improvement in the Hamilton depression (p<0.05). Apparent sadness, reported sadness, and pessimistic scale item 1 (depressed mood) was as early as week 1 (p= thoughts were significantly improved in both quetiapine groups as 0.003) for both quetiapine doses and continued to be sta- early as week 1 compared with placebo (p<0.05). Both doses of quetiapine were approximately twice as effective as placebo in re- tistically superior to placebo at all time points. ducing suicidal thoughts at the final assessment (p≤0.01). b p<0.001 versus placebo. Clinical Global Impression. Quetiapine-treated patients c p<0.01. experienced a statistically significant improvement d < p 0.05. (p<0.001) on the CGI severity scale as early as week 1 that was sustained to the end of the study for both quetiapine ing Scale total score at week 8 in the patients with rapid doses versus placebo. At the final assessment, a larger per- cycling was –17.7 in the 600 mg/day quetiapine group and centage of patients were rated as “normal, not at all ill,” or

1356 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 CALABRESE, KECK, MACFADDEN, ET AL.

TABLE 4. Incidence and Withdrawals Because of Adverse Events Occurring in at Least 10% of the Patients in Any Group of Outpatients with Bipolar I or II Disorder Who Experienced a Major Depressive Episode Patients Taking 600 mg/day Patients Taking 300 mg/day of Quetiapine of Quetiapine Patients Taking Placebo (N=180) (N=179) (N=180) Leading to Leading to Leading to Incidence Withdrawal Incidence Withdrawal Incidence Withdrawal Adverse Event N%N%N%N%N%N% Dry mouth 73 40.6a 21.1a 79 44.1a 0 0.0 14 7.8 0 0.0 Sedation 58 32.2a 17 9.4a 53 29.6a 10 5.6a 11 6.1 0 0.0 Somnolence 44 24.4a 52.8a 49 27.4a 73.9a 15 8.3 0 0.0 Dizziness 41 22.8a 63.3a 30 16.8a 10.6a 15 8.3 0 0.0 Fatigue 21 11.7 1 0.6 16 8.9 0 0.0 13 7.2 0 0.0 Constipation 20 11.1a 10.6a 21 11.7a 0 0.0 8 4.4 0 0.0 Headache 18 10.0 1 0.6a 22 12.3 0 0.0 36 20.0 0 0.0 Nausea 6 8.9 0 0.0 14 7.8 3 1.7 23 12.8 0 0.0 Upper respiratory tract infection not otherwise specified 13 7.2 0 0.0 9 5.0 0 0.0 18 10.0 0 0.0 a Significantly higher than placebo (p<0.05).

“borderline ill” in the 600 mg/day (42.4%) and 300 mg/day ment and Satisfaction Questionnaire total scores im- quetiapine groups (38.1%) compared with the placebo proved by 11.71 by the last assessment among patients group (23.7%). treated with 600 mg/day of quetiapine and by 10.77 A larger percentage of patients was also rated as “much” among those treated with 300 mg/day of quetiapine, com- or “very much” improved on the CGI improvement scale pared with 6.44 in the placebo group (p<0.001 for both in the 600 mg/day (55.9%) and 300 mg/day quetiapine quetiapine doses versus placebo). groups (64.0%) compared with the placebo group (34.3%) Safety and Tolerability at the final assessment. Adverse events. Common adverse events (whether or Anxiety symptoms. Mean baseline Hamilton anxiety not considered treatment related) occurred in ≥10% of pa- scale scores were 18.7 (SD=7.3), 18.7 (SD=7.3), and 18.9 tients, and withdrawals due to common adverse events (SD=7.3) in the 600 mg/day, 300 mg/day, and placebo are shown in Table 4. The overall rate of study discontinu- groups, respectively (Table 2). By the study end, the mean ation due to adverse events was 26.1% (N=47) in the 600 Hamilton anxiety scale total score had decreased by –8.75 in mg/day group, 16.0% (N=29) in the 300 mg/day group, and the 600 mg/day group, –8.64 in the 300 mg/day group, and 8.8% (N=16) in the placebo group (Figure 1). There were –5.54 in the placebo group (p<0.001 for both quetiapine no significant differences in the rates of serious adverse doses versus placebo). A significant improvement in the events across treatment groups, and none was treatment Hamilton anxiety scale total scores as early as week 1 related: 5.0% (N=9) in the 600 mg/day group and 3.4% (N= (p<0.05) was maintained to the last assessment (p<0.001 for 6) in the 300 mg/day group compared with 8.9% (N=16) in both quetiapine doses versus placebo). Individual items of the placebo group. Two patients attempted suicide (one in the Hamilton anxiety scale that most differentiated que- each of the active treatment groups), but no suicides or tiapine-treated patients from those who received placebo deaths occurred during the study. included anxious mood, depressed mood, insomnia, geni- tourinary symptoms, and tension. A more detailed analysis The rate of discontinuation due to adverse events in the of the results of the effect of quetiapine on anxiety measures subgroup of patients with bipolar I disorder was 23.3% in this study has been presented in a separate report (35). (N=28) in the 600 mg/day group, 13.1% (N=16) in the 300 mg/day group, and 11.9% (N=14) in the placebo group. Quality of sleep. The quality of sleep improved signifi- The incidence of serious adverse events in the subgroup of cantly among those treated with either dose of quetiapine patients with bipolar I disorder was 5.0% (N=6) in the 600 compared with placebo. The mean improvement in Pitts- mg/day group, 4.2% (N=5) in the 300 mg/day group, and burgh Sleep Quality Index scores from baseline in patients 11.9% (N=14) in the placebo group. treated with 600 mg/day (–5.46) and 300 mg/day (–5.16) of In the subgroup of patients with bipolar II disorder, the quetiapine was significantly greater with both doses rate of discontinuation due to adverse events was 31.7% (p<0.001) than with placebo (–2.94). (N=19) in the 600 mg/day group, 22.0% (N=13) in the 300 Quality of life. Quetiapine-treated patients also experi- mg/day group, and 3.2% (N=2) in the placebo group. The enced statistically significant improvements in quality of incidence of serious adverse events in the subgroup of pa- life during the study, as determined by the change from tients with bipolar II disorder was 5.0% (N=3) in the 600 baseline in the Quality of Life Enjoyment and Satisfaction mg/day group, 1.7% (N=1) in the 300 mg/day group, and Questionnaire total scores. Mean Quality of Life Enjoy- 3.2% (N=2) in the placebo group.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1357 QUETIAPINE FOR BIPOLAR DEPRESSION

The incidence of treatment-emergent mania was low Discussion and not significantly different from placebo at either quetiapine dose: 2.2% with 600 mg/day of quetiapine (Coch- To our knowledge, this is the first randomized, parallel- ran-Mantel-Haenszel, odds ratio=0.57, 95% confidence group, placebo-controlled trial to evaluate the efficacy of interval (CI)=0.17–1.91, p=0.35), 3.9% with 300 mg/day of quetiapine in bipolar depression. It may also be the first quetiapine (Cochran-Mantel-Haenszel, odds ratio=0.97, published large-scale, controlled study to assess the effi- 95% CI=0.35–2.68, p=0.95), and 3.9% with placebo. cacy of any pharmacological treatment in a group of patients with bipolar I or II depression, and one of few stud- The mean Simpson-Angus Rating Scale total score de- ies to examine an antidepressant effect in patients with creased in all three groups from baseline to the final as- rapid cycling. sessment by –0.1, –0.2, and –0.3 in the 600 mg/day and 300 mg/day quetiapine groups and the placebo groups, re- Quetiapine monotherapy has significant antidepressant spectively. There was no statistically significant difference efficacy in a group of patients with bipolar I or II depression based on the primary efficacy analysis (mean change in the number of patients with an increase from baseline in Montgomery-Åsberg Depression Rating Scale total in Simpson-Angus Rating Scale scores between either of score from baseline to last assessment). The magnitude of the quetiapine groups and placebo: 15% (logistic regres- the clinical improvement was substantial and evident sion=0.66, df=3, p<0.08), 9% (logistic regression=0.06, df= from the first assessment (week 1) and at each visit there- 3, p=0.89), and 9% in the 600 mg/day and 300 mg/day que- after. The rates of response and remission and the time to tiapine and placebo groups, respectively. response and remission were significantly improved in the At the last assessment, mean Barnes Rating Scale for quetiapine groups compared with placebo. Compared Drug-Induced Akathisia scores were low and similar in all with placebo, evidence of early and sustained efficacy was groups: 0.3 in the 600 mg/day group, 0.2 in the 300 mg/day observed consistently with both doses of quetiapine and group, and 0.1 in the placebo group. There was no statisti- in all secondary efficacy analyses from week 1 onward. cally significant difference in the number of patients with In the Montgomery-Åsberg Depression Rating Scale an increase from baseline in Barnes Rating Scale for Drug- item analysis, both doses of quetiapine produced a signif- Induced Akathisia score between either of the quetiapine icant and early improvement in all of the core mood groups and placebo: 12% (logistic regression=0.39, df=3, symptoms of depression, including objective and re- p=0.31), 9% (logistic regression=0.06, df=3, p=0.89), and ported sadness, anhedonia, and pessimistic thoughts. No- 9% in the 600 mg/day and 300 mg/day quetiapine and pla- tably, both doses of quetiapine were approximately twice cebo groups, respectively. as effective as placebo in reducing suicidal ideation. These Adverse events considered extrapyramidal symptoms findings provide support for the conclusion that quetia- were present in 8.9% of the 600 mg/day group, 6.7% of the pine has specific antidepressant properties. 300 mg/day group, and 2.2% of the placebo group; discon- In this study, significant antidepressant efficacy was tinuation rates for extrapyramidal symptoms were 2.8%, demonstrated for quetiapine dosed once a day in the 1.1%, and 0.6%, respectively. evening. This has important clinical relevance because Laboratory Results and Vital Signs once-daily dosing has been associated with enhanced medication adherence (36). Dosing at bedtime may also No clinically relevant differences between groups were offer a means of improving tolerability, particularly re- seen in the mean change from baseline for any vital signs, garding somnolence or sedation that are sometimes seen ECGs, hematology, or clinical chemistry parameters. with quetiapine and may help treat the sleep disturbance Patients treated with 600 mg/day of quetiapine experi- that often accompanies bipolar depression. enced a mean weight gain of 1.6 kg by the final assessment Both doses of quetiapine were associated with improve- compared with 1.0 kg in the 300 mg/kg group and 0.2 kg in ments in quality of sleep and quality of life and were effec- the placebo group. At the final assessment, 16 patients tive in patients with a recent history of rapid-cycling bipo- (9.0%) treated with 600 mg/day of quetiapine, 15 patients lar disorder. Exploratory analyses suggest that the clinical (8.5%) treated with 300 mg/day of quetiapine, and three effect of both doses of quetiapine was greater in patients patients (1.7%) who received placebo had a weight gain of with bipolar I disorder than those with bipolar II disorder. ≥ 7% of their baseline measurement. No patients withdrew The most common side effects of quetiapine included from the study because of weight gain. dry mouth, sedation, somnolence, dizziness, and constipa- Mean fasting serum glucose levels at baseline were 86 tion. The most common side effects leading to withdrawal (SD=12), 87 (SD=13), and 87 (SD=15) mg/dl in the 600 mg/ from the study were sedation and somnolence, with most day and 300 mg/day of quetiapine and placebo groups, re- discontinuations occurring within the first week. Of impor- spectively. By the final assessment, the mean change in tance, changes in weight observed in all three groups were fasting serum glucose was 6 mg/dl (SD=17), 3 mg/dl (SD= relatively small and did not result in withdrawal from the 13), and 4 mg/dl (SD=26) in the 600 mg/day and 300 mg/ study. Quetiapine treatment was not associated with treat- day of quetiapine and placebo groups, respectively. ment-emergent mania. The long-term safety of quetiapine

1358 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 CALABRESE, KECK, MACFADDEN, ET AL. is being explored in ongoing bipolar disorder maintenance mania and other disorders—were effective, guidance on studies. However, data from patients with schizophrenia the best dosing for most patients or subgroups of patients does not suggest that unexpected adverse effects during should be assessed in future studies. long-term treatment should be expected (37). In conclusion, this large, randomized, double-blind, pla- Several aspects of the design of this study were innova- cebo-controlled study provides the first pivotal data dem- tive. First, the inclusion of patients with bipolar II disorder onstrating that quetiapine monotherapy is efficacious and into a large-scale study of acute bipolar depression was well tolerated for the acute treatment of bipolar depression novel and enhanced the generalizability of the findings, in a group of patients with bipolar I or II disorder. particularly since there is a higher incidence of bipolar II disorder than bipolar I disorder. The inclusion of patients Acknowledgments with rapid cycling was also innovative and enhanced the The BOLDER Study Group is as follows: Mohammed Alam, Amer- generalizability of the findings to this difficult-to-treat ican Med Research, Oak Brook, Ill.; Valerie Arnold, Clinical Neuro- subgroup. Second, rather than focusing solely on depres- science Solutions, Memphis, Tenn.; Charles Bailey, Clinical Neuro- sive symptoms, this study included sleep quality and science Solutions, Orlando, Fla.; Guy Brannon, Brentwood health-related quality-of-life measures. Sleep-quality as- Research Institute, Shreveport, La.; David Brown, Community Clinical Research, Austin, Tex.; Joseph Calabrese, University Hos- sessments (both patient- and bed-partner-rated) indi- pitals of Cleveland/Case University School of Medicine, Cleve- cated improvements in functioning in addition to symp- land, Ohio; John Carman, Carman Research, Smyrna, Ga.; Andrew tom severity, including several dimensions of sleep quality Cutler, CORE Research, Winter Park, Fla.; Bernadette D’Souza, and daytime dysfunction. The quality-of-life scale pro- Midwest Clinical Research, Dayton, Ohio; Naresh Emmanuel, vided novel information regarding the effect of quetiapine Carolina Clinical Research Services, Columbia, S.C.; Lawrence Ginsberg, Red Oak Psychiatry Associates, Houston, Tex.; Ram on social relationships, living/housing arrangements, Gopalan, Comprehensive Neuroscience of Northern VA, Falls physical health, satisfaction with medication, and global Church, Va.; William Granger, Research Strategies, Inc., Reno, Nev.; satisfaction. Improvements in these measures provide ev- Laszlo Gyulai, University of Pennsylvania Bipolar Disorder, Phila- idence for improved function and overall quality of life in delphia; Howard Hassman, Comprehensive Clinical Research, addition to reduction in the symptoms of the illness. Clementon, N.J.; Saul Helfing, Oregon Center for Clinical Investi- gators, Inc., Lake Oswego, Ore.; George Joseph, Clinical Neuro- Moreover, the inclusion of analyses that quantify the science Solutions, Jacksonville, Fla.; Paul Keck, University of Cin- magnitude of the clinical effect through effect size deter- cinnati, Cincinnati, Ohio; Terrence Ketter, Stanford University minations gives clinicians useful information. Knowing if Bipolar Disorder Clinic, Stanford, Calif.; Arif Khan, Northwest a significant difference is caused by a small clinical effect Medical Research Center, Bellevue, Wash.; Ari Kiev, Social Psychia- try Research Institute, New York, Irving Kolin, Kolin Research (<0.4), a moderately sized clinical effect (0.40–0.79), or a Group, Winter Park, Fla.; James Knutson, BHC Fairfax Hospital, > large clinical effect ( 0.79) has the potential of helping the Kirkland, Wash.; Michael Levy, Behavioral Medical Research of clinician make decisions on how to use a new medication Staten Island, New York; H.E. Logue, Birmingham Psychiatry Phar- (38). The effect sizes reported in the bipolar I depression maceutical Services, Inc., Birmingham, Ala.; David Marks, Opti- study by Tohen et al. (14) were 0.32 with olanzapine mum Health Services, La Mesa, Calif.; Greg Mattingly, St. Charles Psychiatric Association, St. Charles, Mo.; Charles Merideth, Affili- monotherapy and 0.68 with olanzapine-fluoxetine combi- ated Research Institute, San Diego, Calif.; Janice Miller, Clinical nation therapy compared with 1.09 in the bipolar I sub- Neuroscience Solutions, West Palm Beach, Fla.; Dennis Munjack, group with 600 mg/day of quetiapine in this study. Southwestern Research, Inc., Beverly Hills, Calif.; William Privit- This study had several limitations. First, the number of era, Future Search Trials, Austin, Tex.; Fred Reimherr, University of enrolled patients with bipolar II disorder was not suffi- Utah Medical Center, Salt Lake City, Utah; Robert Riesenberg, At- lanta Center for Medical Research, Atlanta, Ga; Leon Rosenberg, cient to draw firm conclusions regarding efficacy in this Center for Emotional Fitness, Moorestown, N.J.; Leon Rubenfaer, subgroup. For this reason, post hoc analyses conducted in Pioneer Pharmaceutical Research, New Baltimore, Mich.; David the bipolar II subgroup included effect size determina- Sack, Comprehensive Neuroscience Inc., Cerritos, Calif.; Abbey tions, which are less affected by sample size than signifi- Strauss, Comprehensive Neuroscience Inc., Boynton Beach, Fla.; cance testing. Second, moderate rates of sedation or som- David Walling, CNS Network, Garden Grove, Calif.; Richard Weisler, Richard H. Weisler M.D. and Associates, Raleigh, N.C. nolence were observed in both quetiapine groups, which might have compromised the integrity of the double-blind Received Aug. 4, 2004; revision received Oct. 28, 2004; accepted design. If this were a significant factor in the assessment of Dec. 10, 2004. From the University Hospitals of Cleveland/Case Uni- efficacy, the reduction in Montgomery-Åsberg Depression versity School of Medicine, Cleveland, Ohio; the Psychopharmacol- ogy Research Program, Department of Psychiatry, University of Cin- Rating Scale total score in patients experiencing sedation cinnati College of Medicine, Cincinnati, Ohio; the Mental Health Care or somnolence would have been greater than those in pa- Line and General Clinical Research Center of the Cincinnati Veterans tients not experiencing these adverse events. However, Affairs Medical Center, Cincinnati, Ohio; AstraZeneca, Wilmington, Del.; the Department of Psychiatry and Behavioral Sciences, Stan- this was not the case, and the improvements observed on ford University, Stanford, Calif.; the Department of Psychiatry and the Montgomery-Åsberg Depression Rating Scale were Behavioral Sciences, Duke University, Raleigh, N.C.; and the Depart- comparable in patients with or without sedation or som- ment of Psychiatry and Behavioral Medicine, University of South Florida, Tampa. Address correspondence and reprint requests to Dr. nolence. Third, although the study indicated that the two Calabrese, 11400 Euclid Ave., Suite 200, Cleveland, OH 44106; doses used—chosen because of their efficacy in bipolar [email protected] (e-mail).

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1359 QUETIAPINE FOR BIPOLAR DEPRESSION

Supported by AstraZeneca Pharmaceuticals (grant 5077US/0049). 18. Ghaemi SN, Goldberg JF, Henry CA, Rosenquist KJ, Ko JY, Hsu DJ, The authors thank Max Brady, D.Phil., and Aruna Seth, Ph.D., Wenze SJ, Wankmuller MM: Quetiapine for rapid-cycling bipo- (PAREXEL MMS) for editorial assistance. Financial support for their as- lar disorder: a long-term follow-up study (abstract). Bipolar Dis- sistance was provided by AstraZeneca. ord 2003; 5(suppl 1):50 19. Post RM, Leverich GS, Altshuler LL, Frye MA, Suppes TM, Keck PE Jr, McElroy SL, Kupka R, Nolen WA, Grunze H, Walden J: An References overview of recent findings of the Stanley Foundation Bipolar 1. Hlastala SA, Frank E, Mallinger AG, Thase ME, Ritenour AM, Network (part I). Bipolar Disord 2003; 5:310–319 Kupfer DJ: Bipolar depression: an underestimated treatment 20. Sajatovic M, Mullen JA, Sweitzer DE: Efficacy of quetiapine and challenge. Depress Anxiety 1997; 5:73–83 risperidone against depressive symptoms in outpatients with 2. Judd LL, Akiskal HS, Schettler PJ, Endicott J, Maser J, Solomon psychosis. J Clin Psychiatry 2002; 63:1156–1163 DA, Leon AC, Rice JA, Keller MB: The long-term natural history 21. Sajatovic M, Brescan DW, Periz DE, DiGiovanni SK, Hattab H, of the weekly symptomatic status of bipolar I disorder. Arch Ray JB, Bingham CR: Quetiapine alone and added to a mood Gen Psychiatry 2002; 59:530–537 stabilizer for serious mood disorders. J Clin Psychiatry 2001; 3. Keller MB, Lavori PW, Coryell W, Andreasen NC, Endicott J, Clay- 62:728–732 ton PJ, Klerman GL, Hirschfeld RM: Differential outcome of 22. Suppes T, McElroy SL, Keck PE, Altshuler L, Frye MA, Grunze H, pure manic, mixed/cycling, and pure depressive episodes in Leverich GS, Nolen WA, Chisholm K, Dennehy EB, Post RM: Use patients with bipolar illness. JAMA 1986; 255:3138–3142 of quetiapine in bipolar disorder: a case series with prospec- 4. Post RM, Denicoff KD, Leverich GS, Altshuler LL, Frye MA, tive evaluation. Int Clin Psychopharmacol 2004; 19:173–174 Suppes TM, Rush AJ, Keck PE Jr, McElroy SL, Luckenbaugh DA, 23. Vieta E, Parramon G, Padrell E, Nieto E, Martinez-Aran A, Cor- Pollio C, Kupka R, Nolen WA: Morbidity in 258 bipolar outpa- bella B, Colom F, Reinares M, Goikolea JM, Torrent C: Quetiap- tients followed for 1 year with daily prospective ratings on the ine in the treatment of rapid cycling bipolar disorder. Bipolar NIMH life chart method. J Clin Psychiatry 2003; 64:680–690 Disord 2002; 4:335–340 5. Judd LL, Akiskal HS, Schettler PJ, Coryell W, Endicott J, Maser JD, 24. Zarate CA Jr, Rothschild A, Fletcher KE, Madrid A, Zapatel J: Clin- Solomon DA, Leon AC, Keller MB: A prospective investigation of ical predictors of acute response with quetiapine in psychotic the natural history of the long-term weekly symptomatic status mood disorders. J Clin Psychiatry 2000; 61:185–189 of bipolar II disorder. Arch Gen Psychiatry 2003; 60:261–269 25. Hamilton M: A rating scale for depression. J Neurol Neurosurg 6. Altshuler LL, Gitlin MJ, Mintz J, Leight KL, Frye MA: Subsyndro- Psychiatry 1960; 23:56–62 mal depression is associated with functional impairment in pa- 26. Young RC, Biggs JT, Ziegler VE, Meyer DA: A rating scale for ma- tients with bipolar disorder. J Clin Psychiatry 2002; 63:807–811 nia: reliability, validity and sensitivity. Br J Psychiatry 1978; 7. Goodwin F, Jamison K: Manic-Depressive Illness. New York, Ox- 133:429–435 ford University Press, 1990 27. Montgomery SA, Åsberg M: A new depression scale designed to 8. Keck PE Jr: The management of acute mania. BMJ 2003; 327: be sensitive to change. Br J Psychiatry 1979; 134:382–389 1002–1003 28. Guy W (ed): ECDEU Assessment Manual for Psychopharmacol- 9. Keck PE Jr, Nelson EB, McElroy SL: Advances in the pharmacologic ogy: Publication ADM 76-338. Washington, DC, US Department treatment of bipolar depression. Biol Psychiatry 2003; 53:671–679 of Health, Education, and Welfare, 1976, pp 218–222 10. American Psychiatric Association: Practice Guideline for the 29. Hamilton M: The assessment of anxiety states by rating. Br J Treatment of Patients With Bipolar Disorder (Revision). Am J Med Psychol 1959; 32:50–55 Psychiatry 2002; 159(April suppl) 30. Buysse DJ, Reynolds CF III, Monk TH, Berman SR, Kupfer DJ: The 11. Goodwin GM (Consensus Group of the British Association for Psy- Pittsburgh Sleep Quality Index: a new instrument for psychiat- chopharmacology): Evidence-based guidelines for treating bipo- ric practice and research. Psychiatry Res 1989; 28:193–213 lar disorder: recommendations from the British Association for 31. Endicott J, Nee J, Harrison W, Blumenthal R: Quality of Life En- Psychopharmacology. J Psychopharmacol 2003; 17:149–173 joyment and Satisfaction Questionnaire: a new measure. Psy- 12. Zornberg GL, Pope HG Jr: Treatment of depression in bipolar chopharmacol Bull 1993; 29:321–326 disorder: new directions for research. J Clin Psychopharmacol 32. Simpson GM, Angus JWS: A rating scale for extrapyramidal side 1993; 13:397–408 effects. Acta Psychiatr Scand Suppl 1970; 212:11–19 13. Calabrese JR, Bowden CL, Sachs GS, Ascher JA, Monaghan E, Rudd GD (Lamictal 602 Study Group): A double-blind placebo- 33. Barnes TRE: A rating scale for drug-induced akathisia. Br J Psy- controlled study of lamotrigine monotherapy in outpatients chiatry 1989; 154:672–676 with bipolar I depression. J Clin Psychiatry 1999; 60:79–88 34. Muller MJ, Himmerich H, Kienzle B, Szegedi A: Differentiating 14. Tohen M, Vieta E, Calabrese J, Ketter TA, Sachs G, Bowden C, Mitch- moderate and severe depression using the Montgomery-Ås- ell PB, Centorrino F, Risser R, Baker RW, Evans AR, Beymer K, Dube berg Depression Rating Scale (MADRS). J Affect Disord 2003; 77: S, Tollefson GD, Breier A: Efficacy of olanzapine and olanzapine- 255–260 fluoxetine combination in the treatment of bipolar I depression. 35. Macfadden W, Calabrese JR, McCoy R, Minkwitz M, Wilson E, Arch Gen Psychiatry 2003; 60:1079–1088; correction, 60:176 Mullen J: Antianxiety effects analysis of quetiapine in bipolar 15. Vieta E, Mullen J, Brecher M, Paulsson B, Jones M: Quetiapine depression, in 2004 Annual Meeting New Research Program monotherapy for mania associated with bipolar disorder: com- and Abstracts. Washington, DC, American Psychiatric Associa- bined analysis of two international, double-blind, randomised, tion, 2004, number 743 placebo-controlled studies. Curr Med Res Opin 2005; 21:923–934 36. Bloom BS: Daily regimen and compliance with treatment (edi- 16. Sachs G, Chengappa KN, Suppes T, Mullen JA, Brecher M, De- torial). BMJ 2001; 323:647 vine NA, Sweitzer DE: Quetiapine with lithium or divalproex for 37. Kasper S, Brecher M, Fitton L, Jones AM: Maintenance of long-term the treatment of bipolar mania: a randomized, double-blind, efficacy and safety of quetiapine in the open-label treatment of placebo-controlled study. Bipolar Disord 2004; 6:213–223 schizophrenia. Int Clin Psychopharmacol 2004; 19:281–289 17. DelBello MP, Schwiers ML, Rosenberg HL, Strakowski SM: A 38. Bowden CL, Davis J, Morris D, Swann A, Calabrese J, Lambert M, double-blind, randomized, placebo-controlled study of que- Goodnick P: Effect size of efficacy measures comparing dival- tiapine as adjunctive treatment for adolescent mania. J Am proex, lithium and placebo in acute mania. Depress Anxiety Acad Child Adolesc Psychiatry 2002; 41:1216–1223 1997; 6:26–30

1360 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Article

Risperidone Treatment of Autistic Disorder: Longer-Term Benefits and Blinded Discontinuation After 6 Months

Research Units on Objective: Risperidone is effective for Results: Part I included 63 children. The Pediatric Psychopharmacology short-term treatment of aggression, tem- mean risperidone dose was 1.96 mg/day per outbursts, and self-injurious behavior at entry and remained stable over 16 Autism Network in children with autism. Because these weeks of open treatment. The change on behaviors may be chronic, there is a need the Aberrant Behavior Checklist irritability to establish the efficacy and safety of subscale was small and clinically insignifi- longer-term treatment with this agent. cant. Reasons for discontinuation of part I included loss of efficacy (N=5) and ad- Method: The authors conducted a mul- verse effects (N=1). The subjects gained tisite, two-part study of risperidone in an average of 5.1 kg. Part II included 32 children ages 5 to 17 years with autism patients. The relapse rates were 62.5% for accompanied by severe tantrums, aggres- gradual placebo substitution and 12.5% sion, and/or self-injurious behavior who for continued risperidone; this difference showed a positive response in an earlier was statistically significant. 8-week trial. Part I consisted of 4-month open-label treatment with risperidone, Conclusions: Risperidone showed per- starting at the established optimal dose; sistent efficacy and good tolerability for part II was an 8-week randomized, dou- intermediate-length treatment of chil- ble-blind, placebo-substitution study of dren with autism characterized by tan- risperidone withdrawal. Primary out- trums, aggression, and/or self-injurious come measures were the Aberrant Be- behavior. Discontinuation after 6 months havior Checklist irritability subscale and was associated with a rapid return of dis- the Clinical Global Impression improve- ruptive and aggressive behavior in most ment scale. subjects.

(Am J Psychiatry 2005; 162:1361–1369)

A utistic disorder is characterized by impaired social finity of medicines such as risperidone for 5-HT recep- interaction, abnormal language development, and repet- tors, especially those of the 5-HT2A and 5-HT2C classes itive and restricted patterns of behavior (DSM-IV), and it (6). Until 2002, only one placebo-controlled study of ris- affects as many as 20 people per 10,000 (1, 2). Children peridone in adults with autism (7) and a handful of open- with autism display broad differences in abilities and label studies of children with pervasive developmental needs, but accompanying maladaptive behaviors such as disorders (8) had been published. In a prior report, we self-injurious behavior, aggression, and tantrums are described the short-term efficacy of risperidone over pla- common and frequently severe enough to limit educa- cebo in an 8-week controlled trial for 101 children and tional and developmental progress. A variety of treat- adolescents with autistic disorder (4). Risperidone was ments, including medication, are employed in the man- chosen for study given the greatest preliminary evidence agement of these maladaptive behaviors. Controlled for its efficacy in this population (6, 8). Although the ef- trials of medication treatment of autism are limited, but fects of risperidone on aggression, tantrums, and self-in- evidence provides support for both conventional and jurious behavior were substantial in our short-term study, atypical antipsychotic agents (3, 4). Among studies of the the question of whether these improvements would en- conventional antipsychotics, well-controlled trials of ha- dure over time remained unanswered. loperidol have revealed statistically significant effects, In this study, subjects who showed a positive response but only modest clinical benefits, in children with autism, to risperidone in the short-term trial were enrolled in an and short- and longer-term side effects are of concern (3, 5). additional 4-month open-label trial (total drug exposure= Atypical antipsychotics appear to be preferred by clinicians 6 months), which was followed by a placebo-controlled because of the perception that atypicals have a more fa- discontinuation protocol lasting up to 8 weeks. The aim of vorable side effect profile than typical neuroleptics, but the study was threefold: first, to determine if the short- few direct comparison data exist. Atypical agents are also term efficacy of risperidone is maintained over time; sec- of interest because of possible serotonin (5-HT) abnor- ond, to determine if the side effect burden of risperidone malities in some individuals with autism and the high af- remains acceptable over an extended treatment period;

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1361 RISPERIDONE AND AUTISM

TABLE 1. Baseline Characteristics of Children With Autism Who Responded to Risperidone in an 8-Week Trial and Did or Did Not Participate in a 4-Month Open-Label Extension Participated in Extension Phase Did Not Participate in Extension Phase Characteristic (N=63) (N=38) Mean SD Mean SD Demographic profile Age (years) 8.6 2.8 9.0 2.5

Na % Na %

Male sex 49 77.8 33 86.8 Tanner stage I or II 55 87.3 33 86.8 Race or ethnic group White 44 69.8 23 60.5 Black 6 9.5 5 13.2 Hispanic 3 4.8 4 10.5 Asian or Pacific Islander 4 6.3 4 10.5 Other (mixed race) 6 9.5 2 5.3 Parental annual income (dollars)b ≤20,000 6 9.5 7 18.4 20,001–40,000 17 27.0 11 28.9 40,001–60,000 15 23.8 2 5.3 >60,000 23 36.5 18 47.4 Parental education Less than high school 1 1.6 2 5.3 High school 13 20.6 6 15.8 Trade or technical school 4 6.3 2 5.3 At least some college 35 55.6 23 60.5 Advanced degree 10 15.9 5 13.2 Living at home with at least one parent 56 88.9 36 94.7 Educational placement Regular class 4 6.3 4 10.5 Regular school, special education 47 74.6 27 71.1 Special school 11 17.5 4 10.5 Residential school 0 0.0 1 2.6 Other 1 1.6 2 5.3 Developmental profile IQ Average or above 3 4.8 2 5.3 Borderline 7 11.1 5 13.2 Mental retardation Mild 17 27.0 8 21.1 Moderate 12 19.0 6 15.8 Severe 11 17.5 7 18.4 Profound 7 11.1 6 15.8 Unable to assess 6 9.5 4 10.5

Mean SD Mean SD Scores on Vineland Adaptive Behavior Scales Communication 42.6 14.6 44.8 17.0 Socialc 46.4 11.9 51.2 15.7 Daily living 38.5 18.7 35.3 18.6 Maladaptive (part 1) 26.4 5.5 25.2 13.2 Clinical profile Scores on Aberrant Behavior Checklist subscales Irritability 26.6 7.5 24.6 6.8 Social withdrawal 15.6 7.7 17.4 9.5 Stereotypy 10.2 4.3 9.0 5.2 Hyperactivityd 33.4 8.5 29.7 9.5 Inappropriate speech 6.0 4.0 5.2 3.7

Na % Na % Clinical Global Impression severity score 4 (moderate) 9 14.3 9 23.7 5 (marked) 36 57.1 19 50.0 6 (severe) 17 27.0 10 26.3 7 (extreme) 1 1.6 0 0.0 Current anticonvulsant treatment 1 1.6 3 7.9 Past treatment profile Medication naive 11 17.7 7 18.4 Antipsychotic 4 6.5 1 2.6 Selective serotonin reuptake inhibitor 9 14.5 7 18.4 Stimulant 16 25.8 5 13.2 Alpha-2 agonist 11 17.7 5 13.2 a The sum of the numbers of subjects may not be the same as the total number because data were missing for some variables. b Nearly significant difference between groups (χ2=6.94, df=3, p<0.08). c Nearly significant difference between groups (t=1.74, df=99, p<0.09). d Significant difference between groups (t=2.03, df=99, p<0.05).

1362 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 RUPP AUTISM NETWORK and third, to examine the feasibility of risperidone discon- included intelligence testing, the Vineland Adaptive Behavior tinuation after 6 months of treatment. Scales (12, 13), routine laboratory tests, an electrocardiogram, measurements of height, weight, and vital signs, a medical history, and a physical examination. All subjects were also required Method to manifest clinically significant problems consisting of aggression, tantrums, and/or self-injury as defined by a score of 18 or Setting and Subjects higher on the irritability subscale of the Aberrant Behavior This study was an extension of the 8-week double-blind, pla- Checklist—Community version (14, 15) rated by the parent (or cebo-controlled trial with parallel groups and the 8-week open- primary caretaker) and confirmed by clinician review. Scores on label risperidone treatment offered to placebo nonresponders. this 15-item subscale range from 0 to 45, with higher scores indi- These short-term trials were conducted by the Autism Network cating greater pathology. In addition, the subjects were required of the National Institute of Mental Health (NIMH) Research to receive a score of moderate or higher on the Clinical Global Units on Pediatric Psychopharmacology between June 1999 and Impression (GCI) severity scale (16, pp. 218–222) from the exam- April 2001. The five clinical sites included the University of Cali- ining clinician. fornia at Los Angeles, Ohio State University, Indiana University, Eligible subjects were randomly assigned to receive risperi- Yale University, and Kennedy Krieger Institute (Johns Hopkins done or placebo for 8 weeks; details are provided elsewhere (4). University). The protocol was approved by the institutional re- The primary outcome measures were the parent-rated irritability view board at each site and the NIMH Data and Safety Monitor- subscale of the Aberrant Behavior Checklist and the clinician- ing Board, and written informed consent was obtained from a rated CGI improvement scale. Subjects showing a 25% reduction parent or guardian prior to enrollment. Safety and protocol on the irritability subscale and a rating of much improved or very fidelity were monitored through weekly conference calls involv- much improved on the CGI improvement scale by the blinded ing the principal investigators and study coordinators, through clinical evaluator were classified as positive responders and were quarterly reviews by the Data and Safety Monitoring Board, and eligible for the 4-month open-label extension phase. At the end of through annual site visits by the coordinating center (Yale the 4-month extension, responders were eligible for the random- University). ized discontinuation phase. In the discontinuation phase, relapse This study enrolled 63 of the 101 subjects in the first protocol, was defined as a 25% increase in the score on the parent-rated Ab- all of whom met criteria for autistic disorder as defined in DSM- errant Behavior Checklist irritability subscale and a CGI improve- IV. Other entry criteria (at the time of enrollment in the initial ment rating of much worse or very much worse, compared to the double-blind phase) included 1) age of 5–17 years, 2) significant prediscontinuation baseline, for at least 2 consecutive weeks, as tantrums, aggression, self-injurious behavior, and/or agitation, 3) assessed by a blinded clinician. absence of significant medical problems and any other psychiat- Secondary parent-rated outcome measures included the four ric disorder requiring drug therapy (e.g., bipolar disorder, psycho- additional subscales of the Aberrant Behavior Checklist: lethargy/ sis), 4) weight of at least 15 kg, and 5) mental age of at least 18 social withdrawal, stereotypic behavior, hyperactivity, and inap- months. No concomitant treatment with psychotropic medica- propriate speech. At baseline, the parents were interviewed to tion was allowed during any phase of the study, except anticon- identify the child’s target symptoms and to rate compulsive be- vulsant treatment for seizure control if the child had been taking haviors on the Children’s Yale-Brown Obsessive Compulsive Scale a stable dose for 4 weeks and had been free of seizures for 6 (see references 4, 13, and 17 for details on study assessment). months (see reference 9 for a detailed description). Medication Dosing Protocol Schedule and Design The medication schedule in the initial 8-week trial was based At the final visit of the initial 8-week controlled study, or the 8- on the child’s weight and clinical response. The maximum risperi- week open-label risperidone treatment for placebo nonre- done dose was 2.5 mg/day for children between 15 and 45 kg and sponders, subjects deemed responders were offered enrollment 3.5 mg/day for children weighing above 45 kg. Dose reductions to in the extension protocol. Written informed consent for contin- manage side effects were allowed at any time. During the 4- ued participation was obtained from the parents, and for 10 sub- month open phase, clinicians were allowed to adjust the total jects deemed to have the capacity to consent, assent was ob- daily dose according to response and/or adverse events, with the tained. The subjects were then seen every 4 weeks for 4 months for assessment of efficacy, safety, and possible need for dose ad- total daily dose increased up to a maximum of 3.5 mg/day for justments. At the end of these 4 months of open-label treatment, children weighing 15–45 kg and up to 4.5 mg/day for children subjects who continued to show a positive response entered the above 45 kg. discontinuation phase. In this phase, the subjects were randomly Safety Monitoring assigned again, this time either to continued risperidone at the same dose or to gradual placebo substitution, in a double-blind The routine laboratory tests, electrocardiogram, and physical fashion. The discontinuation reduced the maintenance dose by examination were repeated at entry into the extension phase and 25% per week. Thus, the dose was 75% of the week 16 dose for 1 prior to the discontinuation. Weight and vital signs were as- week, followed by 50% of the week 16 dose for the second week, sessed at each visit. At each visit, the primary clinician asked the 25% of the week 16 dose for the third week, and placebo only by parents about the child’s health complaints, intercurrent illness, the fourth week. All subjects were seen weekly for a total of 8 and concomitant medications, and the clinician administered a weeks in the discontinuation phase. 32-item checklist concerning the child’s energy level, muscle stiffness, motor restlessness, bowel and bladder habits, sleep, Baseline Assessment and Measurement of Outcome and appetite. Neurological side effects were further assessed at At screening for participation in the initial 8-week treatment each visit with the Simpson-Angus Rating Scale (18) and the Ab- study, autistic disorder was ascertained through a history and ex- normal Involuntary Movements Scale (AIMS) (16, pp. 534–537). amination by a research team and was corroborated by the Au- Adverse events indicated by any of these methods were docu- tism Diagnostic Interview—Revised (10), administered by a clini- mented according to severity, duration, attribution, outcome, cian trained to reliability (11). The screening measures also and action taken.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1363 RISPERIDONE AND AUTISM

TABLE 2. Scores on Subscales of the Aberrant Behavior FIGURE 1. Scores on the Irritability Subscale of the Aber- Checklist for 63 Children With Autism Who Responded to rant Behavior Checklist for 63 Children With Autism Who Risperidone in an 8-Week Trial and Participated in a 4- Responded to Risperidone in an 8-Week Trial and Partici- Month Open-Label Extension pated in a 4-Month Open-Label Extension Analysis Week 0a Endpointb (df=1, 58) 16 Subscale Mean SD Mean SD F p Irritability 9.5 6.8 11.7 8.0 5.35 0.02 Lethargy/social 14 withdrawal 7.3 5.4 6.8 5.9 2.61 n.s. Stereotypy 4.9 4.3 5.8 4.7 5.47 0.02 Hyperactivity 15.1 10.0 15.8 10.2 0.43 n.s. Inappropriate 12 speech 3.4 3.6 3.4 3.2 0.39 n.s. a Week 0 corresponds to the end of the initial 8 weeks of medication exposure. b For patients who discontinued treatment, the last observation was 10 carried forward.

of Aberrant Behavior Checklist 8 Statistical Analysis on IrritabilityMean Score Subscale Open-label trial. The statistical tests were two-tailed with alpha set at 0.05. To evaluate differences between subjects who participated in the extension trial and those who did not, we compared 6 demographic and clinical characteristics by t tests for continuous 0a 4 8 12 16 Endpointb variables and chi-square tests for categorical variables. All sub- Week jects who enrolled in the extension trial were included in an intent-to-treat analysis. The monthly scores on the parent-rated a Week 0 corresponds to the end of the initial 8 weeks of medication subscales (irritability, lethargy, stereotypy, hyperactivity, and in- exposure. b appropriate speech) of the Aberrant Behavior Checklist were ana- For patients who discontinued treatment, the last observation was carried forward. lyzed with mixed effects linear regression models in which time, site, and site-by-time interaction were the fixed effects (19). The average slope of the regression line (change in irritability score or the efficacy analysis. During the data analysis, three sub- score on other subscale) was calculated for each individual across jects were noted to have entered the extension phase time. In addition, the distribution of CGI improvement scores was calculated across time. without fully meeting the response criteria. Inclusion of these three subjects did not alter the results, and thus Discontinuation phase. As originally planned in the protocol, interim analyses were performed for the Data and Safety Moni- they were included in all analyses. Demographic charac- toring Board after the first 16 and 32 subjects completed this teristics of the entire 63 subjects in the extension study phase of the study (9). These interim analyses were reviewed only are shown in Table 1. When compared to the 38 subjects by the Data and Safety Monitoring Board, independently of the who did not participate in the extension phase, the ex- study investigators. Chi-square analysis with Yates’s correction tension group showed few differences. Of a total of 24 was used to compare the rates of relapse in the two groups. To evaluate the time to relapse and the probability of relapse, the log comparisons, only one variable, the baseline score on the rank survival analysis of continued drug versus placebo and the Aberrant Behavior Checklist hyperactivity subscale, dif- Kaplan-Meier procedure were used. fered significantly, being slightly higher at baseline in the subjects continuing in the extension study. Results 4-Month Open-Label Treatment: Outcomes Baseline Characteristics A total of 51 subjects (81.0%) completed the 16-week Of the original 101 subjects (82 boys and 19 girls) en- open-label treatment period. Of the 12 dropouts from the rolled in the short-term trial, 63 subjects (mean age=8.6 extension study, five subjects were withdrawn because of years, SD=2.8) showed a positive response to risperidone loss of efficacy, one was withdrawn because of noncompli- and consented to participate in the extension protocol. ance with the protocol, one withdrew because of an ad- Of these 63 subjects, 30 were classified as responders verse event (constipation), one withdrew consent (no during the double-blind trial and 30 were so classified in longer interested in medication treatment), and four were the 8-week open-label risperidone trial for placebo non- lost to follow-up. The mixed effects regression model for responders. There were no differences in scores on the ir- all 63 subjects in the intent-to-treat analysis revealed a sig- ritability subscale of the Aberrant Behavior Checklist or nificant time effect for the irritability subscale of the Aber- in the distribution of CGI severity scores at the baseline rant Behavior Checklist (Table 2, Figure 1). However, the of the extension phase between subjects who entered di- mean irritability ratings showed only a minor increase rectly from the double-blind study and those who en- across the 16-week extension phase, from a mean score of tered from the 8-week open-label trial for placebo nonre- 9.5 (SD=6.8) at the start to 10.8 (SD=7.1) at week 16 (Figure sponders. Therefore, the two groups were combined in 1), in contrast to the mean pretreatment irritability score

1364 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 RUPP AUTISM NETWORK

TABLE 3. Clinical Global Impression (CGI) Improvement Ratings for 63 Children With Autism Who Responded to Risperi- done in an 8-Week Trial and Participated in a 4-Month Open-Label Extension Week 0a Week 4 Week 8 Week 12 Week 16 Endpoint Rating N% N% N% N% N% N% 1 (very much improved) 19 30.2 13 20.6 20 33.9 15 26.3 17 33.3 19 30.2 2 (much improved) 42 66.7 38 60.3 31 52.5 28 49.1 29 56.9 33 52.4 3 (minimally improved) 0 0.0 7 11.1 5 8.5 11 19.3 2 3.9 6 9.5 4 (no change) 2 3.2 1 1.6 1 1.7 1 1.8 0 0.0 0 0.0 5 (worse) 0 0.0 3 4.8 1 1.7 2 3.5 2 3.9 3 4.8 6 (much worse) 00.011.611.700.012.023.2 Total 63 100.0 63 100.0 59 100.0 57 100.0 51 100.0 63 100.0 a Week 0 corresponds to the end of the initial 8 weeks of medication exposure. of 26.6 (SD=7). This 1-point increase lacks clinical signifi- tion phase and the remainder of the 63 subjects, aside cance. Indeed, after 16 weeks of treatment the mean score from the discontinuation subjects’ again showing higher still showed a 59% reduction from the mean rating at the mean baseline scores on the Aberrant Behavior Checklist initiation of risperidone treatment 6 months before, a per- hyperactivity factor (mean=34.4, SD=8.7, versus mean= cent reduction identical to that seen after the initial 8- 30.6, SD=9.0; t=2.08, df=99, p=0.04). Also, subjects in the week double-blind study and yielding an effect size of discontinuation phase showed higher mean baseline d>1.0. Analysis of the CGI improvement ratings at end- scores on the Aberrant Behavior Checklist irritability factor point showed that 82.5% of the subjects (N=52) continued than the remaining 63 subjects (mean=27.6, SD=6.1, versus to be rated as much improved or very much improved (1 mean 24.8, SD=7.7, t=2.02, df=99, p=0.05). As planned, the or 2 on the CGI improvement scale) (Table 3). By contrast, NIMH Data and Safety Monitoring Board independently only 7.9% were rated as worse or much worse compared to reviewed two interim efficacy analyses, first after the ini- baseline. tial 16 subjects completed this phase and again after the The risperidone doses were also examined to determine first 32 subjects finished. The relapse rate after 32 subjects whether a dose increase is required to ensure stability of completed this phase was significantly higher in the pla- treatment effects. The mean risperidone doses were 1.96, cebo-treated group (62.5%, N=10) than in the group re- 1.80, 1.87, 2.10, and 2.08 mg/day for weeks 0, 4, 8, 12, and ceiving continued risperidone (12.5%, N=2) (Yates’s cor- 16, respectively, representing a modest 6% increase over rected χ2=6.53, df=1, p=0.01). The median survival time, the 4-month treatment period. i.e., time to relapse, was 34 days for the placebo-treated Secondary outcome measures included the other sub- subjects versus 57 days for those continuing to take ris- scales of the Aberrant Behavior Checklist. Table 2 shows peridone (Figure 2). On the basis of the results of this scores on those subscales and results of the random re- planned interim analysis, the NIMH Data and Safety gression model. Monitoring Board ruled that the discontinuation phase be stopped immediately. The four subjects who were still 4-Month Open-Label Treatment: Adverse Events in this phase of the protocol exited the study and were Adverse events, especially mild to moderate increased treated clinically. Exploratory post hoc analyses were per- appetite, tiredness, and/or drowsiness, were common (Ta- formed in an effort to identify any clinical predictors of re- ble 4). Only one subject withdrew because of an adverse lapse. Age, initial Aberrant Behavior Checklist irritability event (constipation). Although the parents of six subjects score, and IQ all failed to differ significantly (p>0.30) be- (9.5%) reported “abnormal movements,” no dyskinesias tween the relapsing and nonrelapsing subjects. were identified on repeated examination with the AIMS and Simpson-Angus scale by a physician. One subject with Discussion preexisting obesity had galactorrhea according to the mother’s report, but this was not observed on examina- Data from this study suggest that risperidone is a well- tion. Compared to their weight at the initiation of risperi- tolerated and effective treatment for up to 6 months for done treatment, the subjects showed a 6-month weight in- children with autism complicated by tantrums, aggres- crease of 5.1 kg (SD=3.6) (paired t=7.46, df=31, p<0.001), sion, and self-injury. As measured by the primary indices which was significantly greater (p<0.001) than the amount of response, the CGI improvement scale and the Aberrant expected on the basis of available developmental norms. Behavior Checklist irritability subscale, improvements as- This finding has been reviewed in more detail in a separate sociated with risperidone administration were main- report (20). tained in over 80% of the subjects, with very good tolerability. Furthermore, gradual substitution of placebo for Discontinuation Phase risperidone was associated with a greater return of aggres- A total of 38 subjects were enrolled in the discontinua- sion, temper outbursts, and self-injurious behavior than tion phase. Few differences in baseline characteristics in the subjects who continued to receive risperidone in were observed between the subjects in the discontinua- the discontinuation phase. Taken together, these data

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1365 RISPERIDONE AND AUTISM

TABLE 4. Adverse Events in 63 Children With Autism Who Responded to Risperidone in an 8-Week Trial and Participated in a 4-Month Open-Label Extension

Number of Subjects With Percent of Subjects Subjects With Events Moderate or Severe Event With Moderate or Related to Study Drug Adverse Event During 4-Month Extensiona Moderate SevereSevere Event N % Nasal congestion 7 0 11.1 0 0.0 Appetite increase 5 0 7.9 4 6.3 Coughing 4 0 6.3 0 0.0 Anxiety 3 0 4.8 0 0.0 Difficulty falling asleep 3 0 4.8 0 0.0 Fever 3 0 4.8 0 0.0 Skin irritation 3 0 4.8 0 0.0 Vomiting 3 0 4.8 1 1.6 Accidental injury 2 0 3.2 0 0.0 Constipation 2 0 3.2 1 1.6 Drowsiness/sedation 1 1 3.2 1 1.6 Enuresis 2 0 3.2 0 0.0 Headache 2 0 3.2 0 0.0 Hypersalivation 2 0 3.2 1 1.6 Sore throat 2 0 3.2 0 0.0 Depression 0 1 1.6 1 1.6 Diarrhea 1 0 1.6 0 0.0 Difficulty urinating 1 0 1.6 0 0.0 Dyskinesia 1 0 1.6 1 1.6 Earache 1 0 1.6 0 0.0 Restlessness/agitation 1 0 1.6 1 1.6 Sinus condition 1 0 1.6 0 0.0 Stomach/abdominal discomfort 1 0 1.6 0 0.0 Tiredness/fatigue 1 0 1.6 0 0.0 Tremor 1 0 1.6 1 1.6 Weight gain 1 0 1.6 1 1.6 Muscle rigidity 1 0 1.6 0 0.0 Other 1 0 1.6 0 0.0 a Mild events are not reported. The most severe event is reported if the child reported the event more than once. One subject reportedly experienced galactorrhea (see text). strongly suggest that risperidone is an efficacious treat- In an open-label prospective study of risperidone for 11 ment for short- and intermediate-term management of children with autistic disorder, Zuddas et al. (21) noted serious behavioral problems in children with autism. that 10 of the 11 showed enduring improvement over the As in our study of short-term risperidone efficacy (4), 12-month observation period. Another open trial (22) of risperidone was also associated with continued mainte- risperidone enrolled 22 subjects with autistic disorder. Of nance of improvements on the Aberrant Behavior Check- these, 15 subjects showed significant improvement at a list subscales for hyperactivity, stereotypic behavior, and mean dose of 1.2 mg/day for up to 7 months of treatment. Additional published observations include positive re- lethargy/social withdrawal. In addition, the mean reduc- sponses maintained over a 4-month period in responders tion from baseline in the irritability subscale scores of 59% (23) and improvement maintained during a 2-year treat- at the last observation in the extension phase was nearly ment history in two adults with autistic disorder (24). It identical to that observed in the risperidone group at the should be noted that few earlier studies defined specific completion of the 8-week double-blind efficacy trial. entry criteria for aggression or disruptive behavior, but, These data expand the limited published database on taken together, the available accounts suggest that risperi- extended treatment of autistic disorder with medication. done’s benefit for aggression, severe tantrums, and self-in- Heretofore, the largest extended study of neuroleptic jury accompanying autistic disorder persisted over these treatment for autistic disorder was the examination of the intermediate treatment periods (6–12 months), although effects of haloperidol over a 6-month period. In that study additional long-term treatment data focusing on manag- (5), 60 children treated with a mean dose of 1.23 mg/day of ing severe and challenging behaviors are clearly needed. haloperidol were classified as “good responders.” After 6 A key clinical question concerns the length of continued months of treatment, 33 (63%) of 52 subjects were rated as treatment with risperidone for this clinical indication. On “much improved.” Autism factor ratings derived from the one hand, the return of aggression, tantrums, and agita- Children’s Psychiatric Rating Scale obtained on entry and tion was five times as great in the placebo-substitution at 6 months showed an average decline of 23% from base- group as in the subjects who continued to take risperi- line. Upon abrupt withdrawal of haloperidol, only 14% of done. On the other hand, 37.5% (six of 16) of the children the children were rated as much worse and 50% showed at in the placebo-substitution group did not relapse during least mild deterioration. the 2-month discontinuation period, demonstrating some

1366 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 RUPP AUTISM NETWORK degree of variability in outcome. It is conceivable that FIGURE 2. Survival Analysis for Children With Autism Who more gradual drug tapering may have moderated the ob- Responded to Risperidone and Were Then Randomly As- signed to Placebo or Continued Risperidone for 8 Weeksa served relapse in the risperidone group or that dose reduction, rather than complete medication discontinuation, Taper Placebo (N=16) complete may have been sufficient to maintain treatment gains. It is Risperidone (N=16) also conceivable that concomitant behavioral treatment 100 reinforcing replacement behaviors could protect against relapse following risperidone withdrawal. Nevertheless, 90 the high rate of relapse observed in our study suggests caution when withdrawal of effective treatment for these 80 target symptoms is considered. At a minimum, clinicians should document a clear period of stable functioning be- 70 fore initiating medication taper (25) and ensure that appropriate psychosocial supports are in place to minimize relapse risk. No factors predicting relapse were identifi- 60 able given the modest number of subjects in the discontinuation phase of this study, but they certainly form an 50 important future research question. Longer-term follow- Stable Remaining Percent up information on outcome may help clinicians make de- 40 cisions about maintenance treatment. To this end, we are in the process of obtaining 18-month safety data for our 30 study group. Although adverse events associated with risperidone 10 20 3040 50 60 70 exposure were observed to be common in this study, most Days in Randomized Discontinuation Phase were not deemed clinically significant, and many were not a For the placebo group, the risperidone dose was decreased by 25% attributed to risperidone. It is important that only one per week. Relapse was defined as a 25% increase in the score on the Aberrant Behavior Checklist irritability subscale and a CGI improve- subject withdrew from the 4-month open-label treatment ment rating of much worse or very much worse, compared to the because of an adverse event, and no cases of dyskinesia prediscontinuation baseline, for at least 2 consecutive weeks. The were observed during 6 months of treatment or upon difference between groups was significant (test for equality of survival distributions for treatment, log rank=6.89, df=1, p=0.009). withdrawal. The absence of dyskinesia in this study is noteworthy given the report by Campbell and colleagues (26) that 30% of 118 subjects showed dyskinesia following doses in several other studies (16, 22, 27, 28). The possibil- withdrawal of haloperidol after a similar duration of treat- ity of solidifying these gains or even extending the benefit ment, although the more gradual taper used in this study of risperidone treatment by combining it with behavior may be responsible for the finding. Weight gain associated therapy was not explored in this investigation but is an im- with treatment was significant and in some, but not all, portant research question for future studies. Finally, the cases posed a concern (20), especially since antipsychotic- safety observations of the study were limited to a maxi- related weight gain has been associated with diabetes and mum of 8 months of risperidone exposure in a relatively other adverse outcomes. Longer-term observations to de- small study group. Thus, our data may be insufficient to termine the clinical significance of weight gain as well as estimate precisely the long-term risks of risperidone in other adverse events are needed to evaluate the risk-bene- children. fit ratio for risperidone treatment in children with autistic In summary, intermediate-length treatment with ris- disorder. peridone appeared to be associated with the maintenance There were several limitations in this investigation. of reductions in aggression, agitation, self-injury, and irri- First, although the data collection period in this phase of tability from short-term treatment. There was little evi- the study spanned up to a maximum total of 8 months of dence of accommodation to the effects of risperidone, and risperidone exposure, this does not completely mimic the medication appeared to be well tolerated in a group of clinical practice, as many patients receive treatments like children and adolescents with autistic disorder. Additional risperidone for longer time periods. Also, there was no sys- studies on predictors of stable improvement, longer-term tematic effort to reduce or constrain individual daily doses safety, and approaches combining other interventions are over time, leaving some uncertainty about the lowest pos- of interest. sible long-term dose. This limitation notwithstanding, the relatively low mean dose (approximately 2.0 mg/day) of Acknowledgments risperidone used in this study was effective in managing The following are the authors of this report listed by role and the specific target symptoms of aggression, agitation, and study site: Ohio State University, principal investigator Michael G. self-injury. This mean dose was remarkably similar to the Aman, Ph.D., co-investigators L. Eugene Arnold, M.Ed., M.D.,

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1367 RISPERIDONE AND AUTISM

Ronald Lindsay, M.D., Patricia Nash, M.D., Jill Hollway, B.A.; Uni- 6. McDougle CJ, Scahill L, McCracken JT, Aman MG, Tierney E, Ar- versity of California at Los Angeles, principal investigator James T. nold LE, Freeman BJ, Martin A, McGough JJ, Cronin P, Posey DJ, McCracken, M.D., co-investigators Bhavik Shah, M.D., James Mc- Riddle MA, Ritz L, Swiezy NB, Vitiello B, Volkmar FR, Votolato Gough, M.D., Pegeen Cronin, Ph.D., Lisa Lee, B.A.; Indiana Uni- NA, Walson P: Research Units on Pediatric Psychopharmacol- versity, principal investigator Christopher J. McDougle, M.D., co- ogy (RUPP) Autism Network: background and rationale for an investigators David Posey, M.D., Naomi Swiezy, Ph.D., Arlene initial controlled study of risperidone. Child Adolesc Psychiatr Kohn, B.A.; Yale University, principal investigator, Lawrence Sca- Clin N Am 2000; 9:201–224 hill, M.S.N., Ph.D., co-investigators Andres Martin, M.D., Kath- 7. McDougle CJ, Holmes JP, Carlson DC, Pelton GH, Cohen DJ, Price leen Koenig, M.S.N., Fred Volkmar, M.D., Deirdre Carroll, M.S.N., LH: A double-blind, placebo-controlled study of risperidone in Allison Lancor, B.S.; Kennedy Krieger Institute, principal investi- adults with autistic disorder and other pervasive developmen- gator Elaine Tierney, M.D., co-investigators Jaswinder Ghuman, tal disorders. Arch Gen Psychiatry 1998; 55:633–641 M.D., Nilda Gonzalez, M.D., Marco Grados, M.D.; National Insti- 8. Fisman S, Steele M: Use of risperidone in pervasive develop- tute of Mental Health, principal investigator Benedetto Vitiello, mental disorders: a case series. J Child Adolesc Psychopharma- M.D., co-investigator Louise Ritz, M.B.A.; Columbia University, col 1996; 6:177–190 statistician Mark Davies, M.P.H.; Nathan Kline Institute, data 9. Scahill L, McCracken J, McDougle CJ, Aman M, Arnold LE, Tier- management James Robinson, M.Ed., Don McMahon, M.S. ney E, Cronin P, Davies M, Ghuman J, Gonzalez N, Koenig K, Lindsay R, Martin A, McGough J, Posey DJ, Swiezy N, Volkmar F, Received Oct. 23, 2003; revision received Sept. 22, 2004; accepted Ritz L, Vitiello B: Methodological issues in designing a multisite Oct. 1, 2004. From the Research Units on Pediatric Psychopharma- trial of risperidone in children and adolescents with autism. J cology Autism Network (individual contribuors are listed in Acknowl- Child Adolesc Psychopharmacol 2001; 11:377–388 edgments). Address correspondence and reprint requests to Dr. Mc- 10. Lord C, Rutter M, Le Couteur A: Autism Diagnostic Interview— Cracken, UCLA Neuropsychiatric Institute, 760 Westwood Plaza, Los Revised: a revised version of a diagnostic interview for caregiv- Angeles, CA 90024-1759; [email protected] (e-mail). ers of individuals with possible pervasive developmental disor- Supported by NIMH contracts N01 MH-70010 (principal investiga- ders. J Autism Dev Disord 1994; 24:659–685 tor: Dr. McCracken), N01 MH-70009 (principal investigator: Dr. Sca- 11. Lord C, Pickles A, McLennan J, Rutter M, Bregman J, Folstein S, hill), N01 MH-70001 (principal investigator: Dr. McDougle), and N01 MH-80011 (principal investigator: Dr. Aman); by NIH Division of Re- Fombonne E, Leboyer M, Minshew N: Diagnosing autism: anal- search Resources General Clinical Research Center grants M01 RR- yses of data from the Autism Diagnostic Interview. J Autism Dev 00750 (to Indiana University), M01 RR-00052 (to Johns Hopkins Uni- Disord 1997; 27:501–517 versity), M01 RR-00034 (to Ohio State University), and M01 RR-06022 12. Carter AS, Volkmar FR, Sparrow SS, Wang JJ, Lord C, Dawson G, (to Yale University); by NIMH grant MH-01805 (to Dr. McCracken); and Fombonne E, Loveland K, Mesibov G, Schopler E: The Vineland by funding from the Korczak Foundation (to Dr. Scahill). Study medi- Adaptive Behavior Scales: supplementary norms for individu- cations were donated by Janssen Pharmaceutica. als with autism. J Autism Dev Disord 1998; 28:287–302 The authors thank the following members of the Autism Network of 13. Arnold LE, Aman MG, Martin A, Collier-Crespin A, Vitiello B, the NIMH Research Units on Pediatric Psychopharmacology Scientific Tierney E, Asarnow R, Bell-Bradshaw F, Freeman BJ, Gates-Ul- Advisory Board: C.T. Gordon, M.D., Joseph DeVeaugh-Geiss, M.D., Hen- anet P, Klin A, McCracken JT, McDougle CJ, McGough JJ, Posey rietta Leonard, M.D., Richard Shader, M.D., and Susan Swedo, M.D., for their contributions; the NIMH Data and Safety Monitoring Board and DJ, Scahill L, Swiezy NB, Ritz L, Volkmar F: Assessment in multi- Phillip Walson, M.D., for their comments and guidance; and Erin site randomized clinical trials of patients with autistic disorder: Kustan and Caifeng Fu for assistance in preparing this report. the Autism RUPP [Research Units on Pediatric Psychopharma- The content of this publication does not necessarily reflect the cology] Network. J Autism Dev Disord 2000; 30:99–111 views or policies of the Department of Health and Human Services, 14. Aman MG, Singh NN: Aberrant Behavior Checklist—Commu- nor does mention of trade names, commercial products, or organi- nity: Supplementary Manual. East Aurora, NY, Slosson Educa- zations imply endorsement by the U.S. government. tional Publications, 1994 15. Brown EC, Aman MG, Havercamp SM: Factor analysis and norms for parent ratings on the Aberrant Behavior Checklist— References Community for young people in special education. Res Dev Disabil 2002; 23:45–60 1. Chakrabarti S, Fombonne E: Pervasive developmental disor- 16. Guy W (ed): ECDEU Assessment Manual for Psychopharmacol- ders in preschool children. JAMA 2001; 285:3093–3099 ogy: Publication ADM 76–338. Washington, DC, US Depart- 2. Rutter M, Silberg J, O’Connor T, Simonoff E: Genetics and child ment of Health, Education, and Welfare, 1976 psychiatry, II: empirical research findings. J Child Psychiatry 17. Arnold LE, Vitiello B, McDougle C, Scahill L, Shah B, Gonzalez 1999; 40:19–55 NM, Chuang S, Davies M, Hollway J, Aman MG, Cronin P, Koenig 3. Anderson LT, Campbell M, Adams P, Small AM, Perry R, Shell J: K, Kohn AE, McMahon DJ, Tierney E: Parent-defined target The effects of haloperidol on discrimination learning and be- symptoms respond to risperidone in RUPP Autism Network havioral symptoms in autistic children. J Autism Dev Disord Study: a customer approach to clinical trials. J Am Acad Child 1989; 19:227–239 Adolesc Psychiatry 2003; 42:1443–1450 4. McCracken JT, McGough J, Shah B, Cronin P, Hong D, Aman MG, 18. Simpson GM, Angus JWS: A rating scale for extrapyramidal side Arnold LE, Lindsay R, Nash P, Hollway J, McDougle CJ, Posey D, effects. Acta Psychiatr Scand Suppl 1970; 212:11–19 Swiezy N, Kohn A, Scahill L, Martin A, Koenig K, Volkmar F, Car- 19. Gibbons RD, Hedeker D, Elkin I, Waternaux C, Kraemer HC, roll D, Lancor A, Tierney E, Ghuman J, Gonzalez NM, Grados M, Greenhouse JB, Shea MT, Imber SD, Sotsky SM, Watkins JT: Vitiello B, Ritz L, Davies M, Robinson J, McMahon D (Research Some conceptual and statistical issues in analysis of longitudi- Units on Pediatric Psychopharmacology Autism Network): a nal psychiatric data: application to the NIMH Treatment of De- double-blind, placebo-controlled trial of risperidone in chil- pression Collaborative Research Program dataset. Arch Gen dren with autistic disorder. N Engl J Med 2002; 347:314–321 Psychiatry 1993; 50:739–750 5. Perry R, Campbell M, Adams P, Lynch N, Spencer EK, Curren EL, 20. Martin A, Scahill L, Anderson GM, Aman M, Arnold LE, Mc- Overall JE: Long-term efficacy of haloperidol in autistic chil- Cracken J, McDougle CJ, Tierney E, Chuang S, Vitiello B, Re- dren: continuous versus discontinuous drug administration. J search Units on Pediatric Psychopharmacology Autism Net- Am Acad Child Adolesc Psychiatry 1989; 28:87–92 work: Weight and leptin changes among risperidone-treated

1368 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 RUPP AUTISM NETWORK

youths with autism: 6-month prospective data. Am J Psychiatry 25. Pappadopulos E, Macintyre JC, Crismon ML, Findling RL, Ma- 2004; 161:1125–1127 lone RP, Derivan A, Schooler N, Sikich L, Greenhill L, Schur SB, 21. Zuddas A, Di Martino A, Muglia P, Cichetti C: Long-term risperi- Felton CJ, Kranzler H, Rube DM, Sverd J, Finnerty M, Ketner S, done for pervasive developmental disorder: efficacy, tolerabil- Siennick SE, Jensen PS: Treatment recommendations for the ity, and discontinuation. J Child Adolesc Psychopharmacol use of antipsychotics for aggressive youth (TRAAY), part II. J Am 2000; 10:79–90 Acad Child Adolesc Psychiatry 2003; 42:145–161 22. Malone RP, Maislin G, Choudhury MS, Gifford C, Delaney MA: 26. Campbell M, Armenteros JL, Malone RP, Adams PB, Eisenberg Risperidone treatment in children and adolescents with au- ZW, Overall JE: Neuroleptic-related dyskinesias in autistic chil- tism: short- and long-term safety and effectiveness. J Am Acad dren: a prospective, longitudinal study. J Am Acad Child Ado- Child Adolesc Psychiatry 2002; 41:140–147 lesc Psychiatry 1997; 36:835–843 23. Masi G, Cosenza A, Mucci M, Brovedani P: Open trial of risperidone in 24 young children with pervasive developmental dis- 27. Findling RL, Maxwell K, Wiznitzer M: An open clinical trial of ris- orders. J Am Acad Child Adolesc Psychiatry 2001; 40:1206– peridone monotherapy in young children with autistic disor- 1214 der. Psychopharmacol Bull 1997; 33:155–159 24. Dartnall NA, Holmes JP, Morgan SN, McDougle CJ: Brief report: 28. Buitelaar JK, van der Gaag RJ, Cohen-Kettenis P, Melman CT: A two-year control of behavioral symptoms with risperidone in randomized controlled trial of risperidone in the treatment of two profoundly retarded adults with autism. J Autism Dev Dis- aggression in hospitalized adolescents with subaverage cogni- ord 1999; 29:87–91 tive abilities. J Clin Psychiatry 2001; 62:239–248

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1369 Brief Report

Generalizability of Antidepressant Efficacy Trials: Differences Between Depressed Psychiatric Outpatients Who Would or Would Not Qualify for an Efficacy Trial

Mark Zimmerman, M.D. 289 whose symptom severity was too mild to qualify for an an- Iwona Chelminski, Ph.D. tidepressant efficacy trial, and 187 who would be excluded because they were suicidal or had a comorbid anxiety or sub- Michael A. Posternak, M.D. stance use disorder.

Objective: In recent years the generalizability of antidepres- Results: Compared with patients who would qualify for an an- sant efficacy trials has been questioned. Central to the question tidepressant efficacy trial, patients who would be excluded be- of generalizability is whether there are differences in clinical, cause of comorbidity or suicidality were a more chronically ill demographic, and psychosocial characteristics between pa- group with more previous episodes, greater psychosocial im- tients who would qualify for an antidepressant efficacy trial and pairment, and more personality pathology. those who would not qualify. Conclusions: These findings support further caution in gener- Method: The authors compared three groups: 123 depressed alizing the results from antidepressant efficacy trials to clinical patients who would qualify for an antidepressant efficacy trial, populations.

(Am J Psychiatry 2005; 162:1370–1372)

During the past few years there have been increasing sode, age at onset, lifetime history of suicide attempts, and num- concerns about the generalizability of the results of tightly ber of hospitalizations. Personality disorders were assessed in 391 controlled efficacy trials to real-world clinical practice (1). of the patients with the Structured Interview for DSM-IV Person- ality (5). The Rhode Island Hospital Institutional Review Board We recently applied typically used exclusion criteria to a approved the research protocol, and all patients provided in- large group of depressed outpatients and found that most formed consent. The training of the raters and the reliability of depressed patients treated in routine clinical practice the diagnostic assessments in the MIDAS project have been de- would not have qualified for an antidepressant efficacy scribed in detail elsewhere (6). trial (2). In the present report from the Rhode Island Meth- The present report is based on nonpsychotic patients because treatment recommendations differ for psychotic depression. The ods to Improve Diagnostic Assessment and Services four criteria used to define the excluded group are those used in (MIDAS) project, we turn to the question of the compara- the majority of antidepressant efficacy trials (minimum severity bility of the patients who would or would not qualify for an on the Hamilton depression scale, substantial suicide risk, recent antidepressant efficacy trial. Central to the question of DSM-IV diagnosis of alcohol or drug abuse or dependence, and generalizability is whether there are important differences presence of a comorbid axis I disorder) (7). The comorbid disorders used as the basis for exclusion were current posttraumatic in characteristics that are not the basis for exclusion be- stress disorder (PTSD), panic disorder, generalized anxiety disor- tween patients who would or would not qualify for an an- der, or obsessive-compulsive disorder (OCD). tidepressant efficacy trial. We decided a priori to subdivide the excluded group into those who would be excluded because of low symptom severity and Method those who would be excluded because of comorbidity or suicidality, because these variables have different prognostic implications The study group included 599 psychiatric outpatients 18 years (8). We conducted two planned two-group comparisons of the in- old or older whose principal diagnosis was DSM-IV nonbipolar, cluded patients with each of the excluded groups. A comparison of nonpsychotic major depressive disorder. The majority were the two excluded groups was not conducted because it was pe- women (65.4% [N=392]); their mean age was 39.2 years (SD=12.4). ripheral to our central question of whether patients who are in- On presentation for outpatient treatment, all patients were included or excluded from an antidepressant efficacy trial differ. terviewed by a trained diagnostic rater who administered the Structured Clinical Interview for DSM-IV supplemented with Results questions from the Schedule for Affective Disorders and Schizo- phrenia (SADS) (3) assessing the severity of symptoms during the The mean extracted Hamilton depression scale score for week preceding the evaluation. An extracted 21-item Hamilton all 599 patients was 20.3 (SD=5.9). Forty-eight percent of Depression Rating Scale score (4) was derived from the SADS rat- the patients (N=289) scored below 20, the most commonly ings. The interview also included items from the SADS on best used severity threshold for inclusion in an efficacy trial (9). level of social functioning during the past 5 years, social function- Twelve percent of the patients (N=74) had evidence of a ing during adolescence, and the amount of time employed during the past 5 years. For patients with major depressive disorder we current drug or alcohol use disorder. Thirty-nine patients determined the number of episodes, duration of the current epi- (6.5%) were rated 4 or higher on the SADS suicidal ide-

1370 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BRIEF REPORTS

TABLE 1. Demographic and Psychosocial Characteristics of Depressed Outpatients Who Would Be Included or Excluded From Antidepressant Efficacy Trials Excluded Analysis Suicidal Ideation Included Versus Included Versus Included Low Severity or Comorbidity Excluded for Excluded for Suicidal Characteristic (N=123) (N=289) (N=187) Low Severity Ideation or Comorbidity N % N % N % χ2 p χ2 p

Gender 0.2 n.s. 0.8 n.s. Male 42 34.1 92 31.8 73 39.0 Female 81 65.9 197 68.2 114 61.0 Race 1.0 n.s. 0.7 n.s. White 108 87.8 263 91.0 158 84.5 Nonwhite 15 12.2 26 9.0 29 15.5 Marital status 3.6 n.s. 4.2 n.s. Married 55 44.7 135 46.7 66 35.3 Living with someone as if married 5 4.1 10 3.5 14 7.5 Widowed 2 1.6 5 1.7 3 1.6 Separated 11 8.9 20 6.9 18 9.6 Divorced 22 17.9 36 12.5 32 17.1 Single 28 22.8 83 28.7 54 28.9 Education 5.6 n.s. 5.6 n.s. Less than high school education 17 13.8 20 6.9 25 13.4 High school graduation 22 17.9 65 22.5 52 27.8 Some college 52 42.3 123 42.6 77 41.2 College graduation 32 26.0 81 28.0 33 17.6 Past social functioning (12–18 years)a 0.2 n.s. 5.5 <0.05 Good or excellent 92 74.8 209 72.3 116 62.0 Fair or worse 3125.27927.37138.0 Current social functioning (past 5 years)a 0.2 n.s. 1.7 n.s. Good or excellent 82 66.7 199 68.9 111 59.4 Fair or worse 4133.38930.87640.6 Time out of work (past 5 years)b 0.2 n.s. 7.9 <0.01 Virtually no time 41 39.4 111 41.9 41 23.6 Substantial absence 63 60.6 154 58.1 133 76.4 Episode duration of 2 years or more 37 30.1 113 39.1 80 42.8 3.0 n.s. 5.1 <0.05 More than two episodes 36 29.3 89 30.8 78 41.7 0.1 n.s. 4.9 <0.05 At least one hospitalization 24 19.5 40 13.8 45 24.3 1.8 n.s. 0.9 n.s. At least one suicide attempt 22 17.9 48 16.6 59 31.6 0.1 n.s. 7.2 <0.01 Personality disorder diagnosisc Any personality disorder 19 31.7 65 31.2 70 56.9 0.0 n.s. 10.3 <0.01 Any cluster A personality disorder 5 8.3 6 2.9 13 10.6 3.6 n.s. 0.2 n.s. Any cluster B personality disorder 5 8.3 11 5.3 35 28.5 0.8 n.s. 9.4 <0.01 Any cluster C personality disorder 14 23.3 40 19.2 48 39.0 0.5 n.s. 4.8 <0.05

Mean SD Mean SD Mean SD t p t p

Age (years) 41.4 13.3 39.3 12.7 37.6 11.3 1.5 n.s. 2.6 <0.01 Age at onset of depression (years) 29.2 14.1 27.2 14.0 24.0 12.7 1.3 n.s. 3.3 <0.01 Number of depressive episodes 2.3 2.0 2.5 2.2 3.1 2.7 0.5 n.s. 2.7 <0.05 Current episode duration (weeks) 172.2 428.1 182.6 372.1 217.0 386.8 0.2 n.s. 1.0 n.s. Global Assessment of Functioning Scale score 50.7 8.1 53.9 7.0 47.2 7.6 4.1 <0.001 3.9 <0.001 a Ratings from Schedule of Affective Disorders and Schizophrenia. b Those not expected to work (i.e., retired, student, housewife, physically ill) were excluded from the analysis; percents are based on N=104, N= 265, and N=174 for included patients, those excluded for low severity, and those excluded for suicidal ideation or comorbidity, respectively. c Personality disorders were assessed in 391 patients; percents are based on N=60, N=208, and N=123 for included patients, those excluded for low severity, and those excluded for suicidal ideation or comorbidity, respectively. ation item, indicating the presence of frequent suicidal tients whose symptom severity was too mild to qualify for thoughts with a plan. Forty-one percent (N=246) had cur- an antidepressant efficacy trial and the 187 patients who rent PTSD, generalized anxiety disorder, OCD, or panic would be excluded because they were suicidal or had a co- disorder. Based on all four criteria, 79.5% (N=476) of the morbid anxiety or substance use disorder (Table 1). The 599 patients would be excluded from an antidepressant ef- only difference between the included patients and the ficacy trial. group excluded for low symptom severity was on the Glo- We compared demographic, psychosocial, and clinical bal Assessment of Functioning Scale (DSM-IV, p. 32), characteristics of the 123 patients who would qualify for which was expected because symptom severity is one of an antidepressant efficacy trial with those of the 289 pa- the components of this scale.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1371 BRIEF REPORTS

Compared with the included patients, the patients who should be interpreted and promoted. Antidepressant effi- would be excluded because of comorbidity or suicidal ide- cacy trials are generally limited to patients with the great- ation had more social impairment, more frequently est likelihood of demonstrating drug-placebo differences. missed work because of psychiatric reasons, were more Yet, antidepressant medications are approved and mar- likely to have an episode duration of greater than 2 years, keted without acknowledging the generalizability issue. It experienced more previous episodes, made more suicide seems reasonable that antidepressants should be ap- attempts, and were more likely to have a cluster B or clus- proved and marketed only for the narrow range of patients ter C personality disorder. in whom they are proven effective. Approval for other groups of depressed patients should be dependent on Discussion demonstration of efficacy, analogous to requiring additional study to obtain an indication for pediatric depres- When applying the exclusion criteria of recently pub- sion or bipolar depression. lished antidepressant efficacy trials (7) to our patients, we found that the majority would not qualify for most antide- Received Feb. 17, 2004; revision received July 19, 2004; accepted pressant efficacy trials. The 79% exclusion rate found in Aug. 11, 2004. From the Department of Psychiatry and Human Behav- the present study is consistent with our previous findings. ior, Brown Medical School, Rhode Island Hospital, Providence. Address correspondence and reprint requests to Dr. Zimmerman, Bayside The present report extends our earlier results by demon- Medical Center, 235 Plain St., Providence, RI 02905; mzimmerman@ strating that the demographic, clinical, and psychosocial lifespan.org (e-mail). profile of patients who would be excluded from an antide- Supported in part by NIMH grants MH-48732 and MH-56404. pressant efficacy trial because of a comorbid anxiety or substance use disorder or substantial suicidal ideation dif- References fers from that of patients who would qualify for a trial. Specifically, we found that these excluded patients are a 1. Wells KB: Treatment research at the crossroads: the scientific interface of clinical trials and effectiveness research. Am J Psy- more chronically ill group, with more previous episodes, chiatry 1999; 156:5–10 greater social and occupational impairment, and more 2. Zimmerman M, Mattia JI, Posternak MA: Are subjects in phar- personality pathology. In contrast, the psychosocial and macological treatment trials of depression representative of clinical characteristics of patients who would be excluded patients in routine clinical practice? Am J Psychiatry 2002; 159: because of insufficient symptom severity were virtually 469–473 3. Endicott J, Spitzer RL: A diagnostic interview: the Schedule for the same as the patients who would qualify for an antide- Affective Disorders and Schizophrenia. Arch Gen Psychiatry pressant efficacy trial. 1978; 35:837–844 Antidepressant efficacy trials tend to exclude two 4. Endicott J, Cohen J, Nee J, Fleiss J, Sarantakos S: Hamilton De- groups of patients: those who are less likely to respond to pression Rating Scale: extracted from regular and change versions of the Schedule for Affective Disorders and Schizophre- treatment because they have substantial personality pa- nia. Arch Gen Psychiatry 1981; 38:98–103 thology and associated greater chronicity and poorer psy- 5. Pfohl B, Blum N, Zimmerman M: Structured Interview for DSM- chosocial functioning and those who are less likely to IV Personality: SIDP-IV. Washington, DC, American Psychiatric demonstrate differences in response between active med- Press, 1997 ication and placebo by virtue of the milder severity of their 6. Zimmerman M: Integrating the assessment methods of researchers in routine clinical practice: the Rhode Island Meth- symptoms. Although there are some studies demonstrat- ods to Improve Diagnostic Assessment and Services (MIDAS) ing the efficacy of antidepressants in chronically and project, in Standardized Evaluation in Clinical Practice, vol 22. mildly depressed patients (8), these are relatively few. We Edited by First M. Washington, DC, American Psychiatric Pub- are unaware of any large-scale placebo-controlled studies lishing, 2003, pp 29–74 that have focused on the efficacy of treating depression in 7. Zimmerman M, Chelminski I, Posternak M: Exclusion criteria used in antidepressant efficacy trials: consistency across stud- patients with personality disorders. ies and representativeness of samples included. J Nerv Ment The results of the present study, added to the findings of Dis 2004; 192:87–94 studies demonstrating that a low percentage of applicants 8. Posternak MA, Zimmerman M, Keitner GI, Miller IW: A reevalu- are accepted into antidepressant efficacy trials and that a ation of the exclusion criteria used in antidepressant efficacy trials. Am J Psychiatry 2002; 159:191–200 low percentage of patients in clinical practice would qual- 9. Zimmerman M, Posternak MA, Chelminski I: Symptom severity ify for an antidepressant efficacy trial, call for a discussion and exclusion from antidepressant efficacy trials. J Clin Psy- regarding how the results of antidepressant efficacy trials chopharmacol 2002; 22:610–614

1372 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Brief Report

Strong Inverse Association Between Height and Suicide in a Large Cohort of Swedish Men: Evidence of Early Life Origins of Suicidal Behavior?

Patrik K.E. Magnusson, Ph.D. Method: The authors conducted a record linkage study of the birth, conscription, mortality, family, and census register data David Gunnell, M.D., Ph.D. of 1,299,177 Swedish men followed from age 18 to a maxi- Per Tynelius, M.S. mum of age 49. Results: There were 3,075 suicides over an average follow-up George Davey Smith, M.D., Ph.D. period of 15 years. There was a strong inverse association be- Finn Rasmussen, M.D., Ph.D. tween height and suicide risk. In fully adjusted models, a 5-cm increase in height was associated with a 9% decrease in suicide risk. Objective: Previous studies have found associations between Conclusions: The strong inverse association between height poor fetal and infant growth and the risk of suicide. The au- and suicide may signify the importance of childhood exposure thors’ goal was to investigate the association between height— in the etiology of adult mental disorder or reflect stigmatization a measure of childhood growth—and suicide risk. or discrimination encountered by short men in their adult lives.

(Am J Psychiatry 2005; 162:1373–1375)

Height is a marker of postnatal development as well 1960. For men born after 1973, information on birth weight, birth as factors, including genes, that influence growth. Associ- length, and gestational age were obtained from the Medical Birth ations of several adult chronic diseases with height are Register. thought to indicate the importance of the postnatal envi- Altogether, complete information was available on all variables described above for 1,299,177 (90%) of the subjects. Individuals ronment on disease risk (1). Few previous prospective excluded due to incomplete information were shorter (mean studies have investigated the association of height with height=178.6 cm) than those included (mean height=179.3 cm) suicide. A Swedish study (2) found some evidence of a (t=37.05, df=1,442,921, p<0.0001) and had a higher suicide rate greater risk among the shortest 10% of men studied, but (age- and birth-year-adjusted hazard ratio=1.15, p<0.01). possible socioeconomic confounding factors were not Suicide deaths occurring between age 18 and 49 years were iden- taken into account. The other study, from South Korea (3), tified by using ICD-8 and ICD-9 codes E950–E959 or ICD-10 codes X60–X84. Associations were also investigated with 1) undeter- documented a twofold greater risk of suicide in the short- mined deaths (ICD-8 and ICD-9 codes E980–E989 or ICD-10 codes est compared with the tallest men, but this association Y10–Y34) and 2) alcohol-related deaths (ICD-8 and ICD-9 codes was attenuated after the authors controlled for socioeco- 291, 303, 571, and E860 or ICD-10 codes F10, K70, T51, X45, X65). nomic factors. A further investigation reported inverse as- We used Cox’s proportional hazards models, with age as the sociations between height and hospital admission for self- time axis, to investigate associations between height and suicide. harm (4). In the current article, we examine the relation- Subjects were censored at their date of death, emigration, or the end of follow-up (Dec. 31, 1999). We categorized subjects into ship between height and suicide in a large cohort of Swed- nine groups on the basis of the number of standard deviations ish men. their height was from the overall mean (<2.0, –2 to <–1.5, –1.5 to <–1.0, –1.0 to <–0.5, –0.5 to <0.5, 0.5 to <1.0, 1.0 to <1.5, 1.5 to 2.0 Method and >2.0 standard deviations). We also estimated hazard ratios for 5-cm increments in height. Swedish-born males born 1950–1981 for whom information on their biological parents was available (N=1,654,668) were identi- Results fied in the Swedish Multi-Generation Register; 1,442,923 (87.2%) of these individuals had a record in the Military Service Conscrip- The mean height of the 1,299,177 conscripts was 1.79 m; tion Register 1968–1999. These records were linked with the Med- ical Birth Register, the Cause of Death Register 1968–1999, and the 711,949 (54.8%) came from white-collar families; and Population and Housing Censuses 1970–2000. 3,075 (0.24%) died by suicide. Taller men had a much Height measurements at age 18–19 years were obtained from lower risk of suicide than shorter men (Table 1). The asso- the Military Service Conscription Register. The following possible ciation was linear. A 5-cm increase in height was associ- confounding factors were assessed: maternal and paternal edu- ated with a 9% (95% confidence interval [CI]=7%–12%) de- cational level (six categories); the highest socioeconomic index of crease in suicide risk. The effect of height changed little either parent (four categories); body mass index; date of birth; and conscription center (six centers). Information on marital sta- after adjustment for parental socioeconomic index or the tus at age 30–35 years was available for a subgroup born before participant’s body mass index.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1373 BRIEF REPORTS

TABLE 1. Four Models of Association Between Suicide and Height Measured in Standard Deviations From the Overall Mean in a Cohort of 1,299,177 Swedish Mena Model 1 Model 2 Model 3 Model 4 Height Hazard Ratio 95% CI Hazard Ratio 95% CI Hazard Ratio 95% CI Hazard Ratio 95% CI Standard deviation from overall mean <–2.0 1.57 1.30–1.91 1.57 1.30–1.91 1.58 1.30–1.91 1.54 1.27–1.86 –2.0 to <–1.5 1.14 0.96–1.36 1.14 0.96–1.36 1.15 0.96–1.37 1.12 0.94–1.34 –1.5 to <–1.0 1.19 1.05–1.35 1.19 1.05–1.35 1.20 1.05–1.36 1.18 1.04–1.34 –1.0 to <–0.5 1.11 1.01–1.23 1.11 1.01–1.23 1.12 1.01–1.23 1.11 1.01–1.22 –0.5 to <0.5b 1111 0.5 to <1.0 0.94 0.84–1.06 0.95 0.84–1.06 0.94 0.84–1.06 0.95 0.85–1.07 1.0 to <1.5 0.86 0.75–0.98 0.86 0.75–0.98 0.85 0.75–0.98 0.86 0.76–0.99 1.5 to <2.0 0.82 0.66–1.02 0.82 0.66–1.02 0.81 0.65–1,01 0.83 0.66–1.03 >2.0 0.71 0.52–0.96 0.71 0.52–0.96 0.70 0.52–0.95 0.72 0.53–0.97 Hazard ratio per 5-cm increase in height 0.90* 0.88–0.93 0.90* 0.88–0.93 0.90* 0.88–0.93 0.91* 0.88–0.93 a Model 1 controlled for age; model 2 controlled for age, year of birth, and conscription test center; model 3 controlled for age, year of birth, conscription test center, and body mass index; model 4 controlled for factors included in model 3 plus parental socioeconomic index and education. b Reference category. *p<0.0001.

To assess whether secular increases in height may ac- alcohol-related mortality, suggesting that substance mis- count for the observed associations, we repeated the anal- use may contribute to the observed patterns. ysis using year of birth as the time axis. The associations A strength of our study is that it includes 79% of all men were unchanged: the hazard ratio per 5-cm change in born in Sweden in 1950–1981. It has two main limitations. height was 0.90 (95% CI=0.88–0.93) (Table 1). We were unable to fully assess the possible influence of Compared with height-suicide associations, height was unemployment, relationship breakdown, or mental ill- more strongly related to undetermined deaths (hazard ra- ness on the height-suicide associations. Our findings cannot be generalized to women or older men. Nevertheless, tio per 5-cm increase in height=0.86, 95% CI=0.81–0.90), in Sweden, the United States, and the United Kingdom, and alcohol-related deaths (hazard ratio=0.77, 95% CI= suicides in 18–49-year-old men account for almost half of 0.71–0.83). all suicides. In the subgroup born 1950–1965 and alive in 1990, who There are several explanations for our findings. First, had had the opportunity to complete their education, the psychological stress and disrupted family life in childhood height-suicide association was only weakly attenuated— impair growth (6) and may increase susceptibility to men- from 0.91 (95% CI=0.86–0.95) to 0.92 (95% CI=0.88–0.97)— tal illness and suicidal behavior in later life (7). Short stat- after we controlled for education. Information on marital ure may be associated with a greater risk of psychosis (8), status at age 30–35 years was available for men born be- which in turn influences suicide risk. However, associa- fore 1960, and the fully adjusted hazard ratio in this sub- tions were not attenuated by excluding subjects with psy- group—0.93 (95% CI=0.88–0.98)—was only weakly attenu- chiatric diagnoses at conscription. Short individuals are ated after adjustment for marital status—0.94 (95% CI= more likely to be in a low social class as adults, indepen- 0.89–0.99). In the subgroup of men with birth weight data, dent of their childhood social class (9). Low social class is the hazard ratio for suicide was unchanged after we con- associated with a greater risk of suicide (10). In a subset of trolled for birth weight, gestational age, and birth length subjects, however, we found that educational level, a (data not shown). marker of socioeconomic position, had little effect on the Exclusion of all conscripts with a psychiatric diagnosis associations. Marriage protects against suicide (11), and at conscription (ICD-8 and ICD-9 codes 290–319) from the short individuals may be less likely to marry than taller fully adjusted analysis did not affect the hazard ratio ones (12). Marital status only weakly confounded the associations. Low weight gain in infancy may also be a risk fac- markedly—0.91 (95% CI=0.89–0.94). The height-suicide tor for suicide in adult life (13). Finally, short children tend association was 0.94 (95% CI=0.87–1.00) among those ex- to have lower levels of intelligence and may suffer stigma- cluded for this reason. tization and discrimination (9, 14).

Discussion Received Feb. 6, 2004; revisions received April 23 and July 15, 2004; accepted Aug. 11, 2004. From the Department of Genetics and We found a twofold higher risk of suicide in short men Pathology, Rudbeck Laboratory, Uppsala University, Sweden; the De- than tall men. The associations do not appear to be attrib- partment of Social Medicine, University of Bristol, U.K.; and the Child and Adolescent Public Health Epidemiology Group, Department of utable to socioeconomic confounding or prenatal influ- Public Health Sciences, Karolinska Institute. Address correspondence ences on growth (5). Stronger associations were seen with and reprint requests to Associate Professor Rasmussen, Department

1374 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BRIEF REPORTS of Public Health Sciences, Karolinska Institute, Norrbacka, SE-171 76, 7. Fergusson DM, Woodward LJ, Horwood LJ: Risk factors and life Stockholm, Sweden; [email protected] (e-mail). processes associated with the onset of suicidal behaviour during adolescence and early adulthood. Psychol Med 2000; 30: 23–39 References 8. Gunnell D, Rasmussen F, Fouskakis D, Tynelius P, Harrison G: Patterns of fetal and childhood growth and the development 1. Gunnell D: Commentary: can adult anthropometry be used as of psychosis in young males: a cohort study. Am J Epidemiol a “biomarker” for prenatal and childhood exposures? Int J Epi- 2003; 158:291–300 demiol 2002; 31:390–394 9. Nyström P: Childhood environment, intergenerational mobil- 2. Allebeck P, Bergh C: Height, body mass index and mortality: do ity, and adult health: evidence from Swedish data. J Epidemiol social factors explain the association? Public Health 1992; 106: Community Health 1992; 46:71–74 375–382 3. Song YM, Smith GD, Sung J: Adult height and cause-specific 10. Drever F, Whitehead M: Health Inequalities. London, Her Maj- mortality: a large prospective study of South Korean men. Am esty’s Stationery Office, 1997 J Epidemiol 2003; 158:479–485 11. Kreitman N: Suicide, age and marital status. Psychol Med 4. Jiang GX, Rasmussen F, Wasserman D: Short stature and poor 1988; 18:121–128 psychological performance: risk factors for attempted suicide 12. Murray JE: Marital protection and marital selection: evidence among Swedish male conscripts. Acta Psychiatr Scand 1999; from a historical-prospective sample of American men. De- 100:433–440 mography 2000; 37:511–521 5. Thompson C, Syddall H, Rodin I, Osmond C, Barker DJ: Birth 13. Barker DJP, Osmond C, Rodin I, Fall CHD, Winter PD: Low weight and the risk of depressive disorder in late life. Br J Psy- weight gain in infancy and suicide in adult life. BMJ 1995; 311: chiatry 2001; 179:450–455 1203 6. Wales JKH, Taitz LS: Patterns of growth in abused children, in 14. Downie AB, Mulligan J, Stratford RJ, Betts PR, Voss LD: Are short Human Growth: Basic and Clinical Aspects. Edited by Hernán- normal children at a disadvantage? the Wessex growth study. dez M. Amsterdam, Excerpta Medica, 1992, pp 151–159 BMJ 1997; 314:97–100

Brief Report

Is Violent Method of Suicide a Behavioral Marker of Lifetime Aggression?

Alexandre Dumais, B.Sc. They used structured clinical assessments and personality trait Alain D. Lesage, M.D., M.Phil. scales in interviews with family members of the deceased. Aleksandra Lalovic, M.Sc. Results: Violent method was associated with a higher level of Monique Séguin, Ph.D. lifetime aggression and a higher level of impulsivity. In addition, Michel Tousignant, Ph.D. violent method was associated with lifetime substance abuse or Nadia Chawky, M.Ps. dependence and psychotic disorders. Controlling for age, sex, substance disorders, and other major psychopathology, the au- Gustavo Turecki, M.D., Ph.D. thors found that lifetime aggression and the interaction between impulsivity and aggressive behavior remained associated Objective: The main purpose of this study was to investigate whether the method of suicide is a valid behavioral marker of a with violent method. lifetime history of aggression. Conclusions: These results support the use of violent method Method: The authors applied the psychological autopsy of suicide as a behavioral marker of a higher level of lifetime method to investigate 310 individuals who committed suicide. impulsive-aggressive behaviors.

(Am J Psychiatry 2005; 162:1375–1378)

Suicide is an important public health problem (1). The lent method and age remains unclear (4–7). Since the early method of suicide does not appear to be randomly distrib- work of Åsberg (8), suicide method has been used in neu- uted. For example, violent method of suicide is more often robiological studies of suicide attempters to select sub- used by males than females (2–4), and violent method is jects who are more likely to have low indexes of serotoner- more common in suicide completers affected by psycho- gic neurotransmission, which in turn tend to correlate sis (5). On the other hand, the relationship between vio- with higher levels of aggression (9). However, the relation-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1375 BRIEF REPORTS of Public Health Sciences, Karolinska Institute, Norrbacka, SE-171 76, 7. Fergusson DM, Woodward LJ, Horwood LJ: Risk factors and life Stockholm, Sweden; [email protected] (e-mail). processes associated with the onset of suicidal behaviour during adolescence and early adulthood. Psychol Med 2000; 30: 23–39 References 8. Gunnell D, Rasmussen F, Fouskakis D, Tynelius P, Harrison G: Patterns of fetal and childhood growth and the development 1. Gunnell D: Commentary: can adult anthropometry be used as of psychosis in young males: a cohort study. Am J Epidemiol a “biomarker” for prenatal and childhood exposures? Int J Epi- 2003; 158:291–300 demiol 2002; 31:390–394 9. Nyström P: Childhood environment, intergenerational mobil- 2. Allebeck P, Bergh C: Height, body mass index and mortality: do ity, and adult health: evidence from Swedish data. J Epidemiol social factors explain the association? Public Health 1992; 106: Community Health 1992; 46:71–74 375–382 3. Song YM, Smith GD, Sung J: Adult height and cause-specific 10. Drever F, Whitehead M: Health Inequalities. London, Her Maj- mortality: a large prospective study of South Korean men. Am esty’s Stationery Office, 1997 J Epidemiol 2003; 158:479–485 11. Kreitman N: Suicide, age and marital status. Psychol Med 4. Jiang GX, Rasmussen F, Wasserman D: Short stature and poor 1988; 18:121–128 psychological performance: risk factors for attempted suicide 12. Murray JE: Marital protection and marital selection: evidence among Swedish male conscripts. Acta Psychiatr Scand 1999; from a historical-prospective sample of American men. De- 100:433–440 mography 2000; 37:511–521 5. Thompson C, Syddall H, Rodin I, Osmond C, Barker DJ: Birth 13. Barker DJP, Osmond C, Rodin I, Fall CHD, Winter PD: Low weight and the risk of depressive disorder in late life. Br J Psy- weight gain in infancy and suicide in adult life. BMJ 1995; 311: chiatry 2001; 179:450–455 1203 6. Wales JKH, Taitz LS: Patterns of growth in abused children, in 14. Downie AB, Mulligan J, Stratford RJ, Betts PR, Voss LD: Are short Human Growth: Basic and Clinical Aspects. Edited by Hernán- normal children at a disadvantage? the Wessex growth study. dez M. Amsterdam, Excerpta Medica, 1992, pp 151–159 BMJ 1997; 314:97–100

Brief Report

Is Violent Method of Suicide a Behavioral Marker of Lifetime Aggression?

(Am J Psychiatry 2005; 162:1375–1378)

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1375 BRIEF REPORTS

TABLE 1. Axis I and Axis II Disorders and Impulsive and Aggressive Behaviors in Individuals Who Committed Suicide by Vi- olent or Nonviolent Methods Subjects With Disorder or Behavior Nonviolent Violent Method of Method of Disorder or Behavior Na Suicide Suicide Analysis N % N % Odds Ratio 95% CI χ2 df p Axis I disorders Last 6 months Anxiety disorderb 156 4 12.9 21 16.8 1.36 0.41–5.91 0.41c Mood disorderd 156 21 67.7 72 57.6 0.65 0.25–1.58 1.06 1 0.30 Alcohol and/or drug abuse or dependence 156 9 29.0 51 40.8 1.68 0.68–4.50 1.45 1 0.23 Psychotic disordere 156 0 0.0 11 8.8 Indefinite 0.05c Lifetime Anxiety disorderb 156 4 12.9 24 19.2 1.60 0.49–6.88 0.42 1 0.41 Mood disorderd 156 14 45.2 50 40.0 0.81 0.34–1.95 0.27 1 0.60 Alcohol and/or drug abuse or dependence 156 8 25.8 64 51.2 3.02 1.18–8.36 6.45 1 0.01 Psychotic disordere 156 0 0.0 11 8.8 Indefinite 0.05c Axis II disordersf Cluster A: schizoid, paranoid 157 1 3.1 6 4.8 1.56 0.18–74.18 0.56c Cluster B: borderline, antisocial, histrionic, narcissistic 157 10 31.3 45 36.0 1.24 0.51–3.20 0.74 1 0.389 Cluster C: dependent, obsessive-compulsive, avoidant, depressive, passive-aggressive 157 5 16.1 18 14.4 0.91 0.29–3.42 0.01 1 0.95

Mean SD Mean SD t df p Impulsive and aggressive behaviors Barratt Impulsivity Scale total score 135 64.01 15.22 69.33 14.49 –1.89 144 0.06 Brown-Goodwin Lifetime History of Aggression total score 124 6.44 5.80 13.10 13.87 –2.97 106 0.004 a Numbers of subjects in analysis vary according to the particular analysis listed in each row because information was not available for all subjects on all measures. The degree of overlap between measures is variable. Percents are based on N=31 or N=32 for subjects who committed suicide by violent methods and N=125 for subjects who committed suicide by nonviolent methods. b Panic disorder, agoraphobia without panic disorder, social phobia, specific phobia, obsessive-compulsive disorder, generalized anxiety disorder, anxiety disorder not otherwise specified. c Fisher’s exact test. d Major depression, depression not otherwise specified. e Schizophrenia, schizoaffective disorder, delusional disorder. ship between use of violent method and history of aggres- axis II personality disorders. Before applying the SCID, we used sive behavior remains inconclusive (10, 11). Moreover, it the Schedule for Affective Disorders and Schizophrenia for has not been investigated in suicide completers. There- School-Age Children (17) modified to include questions assessing personality disorders adapted from the Children’s Depression In- fore, the main purpose of this study was to investigate ventory (18). As reported elsewhere (19), diagnoses obtained us- clinical and behavioral correlates of violent method of sui- ing these two different methods had an excellent concordance cide and to assess whether it represents a behavioral rate. Information collected with these interviews, from the Coro- marker of lifetime aggressive behavior. ner’s notes, and from medical records was used by a panel of clinicians who made consensus DSM-IV diagnoses. Method Two or more interviewers were asked to rate the same subject separately, and kappa coefficients for key diagnoses were excel- Subjects for this study were individuals who had committed lent: kappa=0.96 for major depression, kappa=0.98 for alcohol suicide identified through the Coroner’s Office and collected se- abuse/dependence, kappa=1.0 for drug abuse/dependence, quentially from the Greater Montreal area. Seventy-three percent kappa=1.0 for bipolar disorder, kappa=1.0 for schizophrenia, and of the families of these individuals agreed to participate in our kappa=1.0 for cluster B personality disorders. These data are con- clinical study. After a period averaging 4 months, these families sistent with those previously presented by our group (20) and were contacted again for interviews. This study was approved by probably are a direct result of frequent training sessions to avoid our local institutional review board, and written informed con- drifting between interviewers. sent was obtained from all participating families. The Brown-Goodwin Lifetime History of Aggression (21) and Suicides were classified on the basis of information provided the Barratt Impulsivity Scale (22) were used to assess lifetime ag- by the Coroner according to criteria used in previous studies (2, 5, gressive and impulsive behaviors, respectively. The Temperament 12). Overdoses, poisoning, gas, and drowning were classified as and Character Inventory (23) was used to complement this infor- nonviolent methods of suicide; all other methods were classified mation. Internal consistency estimates were excellent overall: al- as violent. pha=0.88 for the informant version of the Brown-Goodwin Life- Psychiatric diagnoses in individuals who committed suicide time History of Aggression, alpha=0.89 for the Barratt Impulsivity were made by means of the psychological autopsy method. This Scale, and alpha between 0.73 and 0.88 for the Temperament and technique has been well validated and is outlined elsewhere (13– Character Inventory. 15). Psychiatric diagnoses were obtained by using the Structured In addition, we compared information obtained using two dif- Clinical Interview (SCID) (16) for DSM-IV axis I diagnoses and ferent informants for the same deceased individual for personal-

1376 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BRIEF REPORTS ity trait measures and observed no significant differences (p val- Discussion ues between 0.25 and 0.94). Furthermore, in studies with living subjects we compared information obtained with an informant To our knowledge, this study is the first to investigate the and the subject and found no significant differences (p values be- relationship between measures of aggression, impulsivity, tween 0.67 and 0.98). and suicide method in individuals who committed sui- Chi-square, odds ratios, and Fisher’s exact tests were used to cide. Overall, our analysis supports the use of suicide compare categorical variables, and t tests were used in the analy- method as a possible behavioral marker of lifetime impul- sis of continuous variables. Logistic regression was used to obtain adjusted risks. sive aggression. We found that younger age was associated with violent Results method and observed a nonsignificant gender effect. The effect of gender is consistent with previous findings (2, 4, A total of 310 individuals who had committed suicide 6), and its nonsignificance may be a consequence of the were investigated: 36 female subjects and 274 male sub- relatively small number of female subjects included in this jects with a mean age of 39.45 years (SD=13.91). Two hun- study. Although suicide cases in this study were obtained dred forty-two (78.1%) of these individuals used a violent sequentially, our group has been focusing on male suicide method: 166 (68.6%) hanged themselves, 39 (16.1%) used and only more recently has started recruiting females. The a firearm, 12 (5.0%) used laceration, 18 (7.4%) jumped age effects found in this study support the finding by Con- from a height, and seven (2.9%) used one of the following: well et al. (7) that older individuals who commit suicide traffic accident, electrocution, self-immolation, or stran- are more likely to use nonviolent methods. Finally, we ob- gulation with a plastic bag. On the other hand, 68 (21.9%) served that lifetime alcohol and/or drug problems and of the individuals who committed suicide used nonviolent psychotic disorders were associated with use of violent methods. Of these, 36 (52.9%) died intoxicated by gases, 22 methods. In addition, significant correlations between (32.4%) by drug poisoning, and 10 (14.7%) by drowning. lifetime history of substance misuse and impulsive behav- This distribution of suicide methods is consistent with ior (r=0.41, N=135, p=0.001) and aggression (r=0.32, N= that reported for all suicide cases by the Quebec Coroner’s 124, p=0.004) were found. These findings are consistent Office in 2000. The average age was lower in the violent with those of previous studies (24, 25) suggesting that method group (38.20 compared with 43.96 years) (t=2.54, there is probably a relationship between chronic alcohol df=305, p<0.02), and female subjects showed a tendency and/or drug consumption and lifetime aggression and vi- to use a nonviolent method (χ2=3.09, df=1, p<0.08). olence. Nevertheless, when we controlled for the history of substance use, as well as for other positive variables, we Last-6-month prevalence rates of anxiety disorders, found that higher levels of impulsive and aggressive be- mood disorders, and alcohol and/or drug problems havior and their interaction remained significant predic- (abuse/dependence) were comparable between groups. tors of violent methods of suicide. Similarly, lifetime prevalence rates of anxiety disorders Given that the lethality associated with violent methods and mood disorders were comparable between groups, as is considerably higher than that associated with nonvio- were the prevalence rates of cluster A, cluster B, and clus- lent methods (26), a possible hypothesis that could be ter C personality disorders. In contrast, prevalence rates of drawn from our results is that the observed excess of im- lifetime alcohol and/or drug problems and psychotic dis- pulsive-aggressive behavior among individuals who com- orders (6-month and lifetime) were significantly associ- mitted suicide may be a direct consequence of the fact ated with violent method of suicide (Table 1). that these individuals were more likely to use a violent Measures of lifetime history of aggressive behaviors method. This hypothesis is supported by recent data sug- were higher in the group that used a violent method of sui- gesting that differences in method explain a substantial cide. In addition, we found a nonsignificantly higher level amount of the difference between completed suicide and of impulsive behaviors as measured by the Barratt Impul- medically serious nonfatal suicide attempts (27). sivity Scale in the violent method group. Aggressive and The major limitation of this study is intrinsic to the psy- impulsive behavior scores were also found to be signifi- chological autopsy method. However, our results on valid- cantly correlated (r=0.45, N=124, p<0.001). On the other ity are consistent with the literature (13–15) and support hand, Temperament and Character Inventory measures the use of behavioral assessments obtained by means of were not statistically different between groups (data not informants. Other limitations in our analyses include the shown). Finally, when we controlled for the effect of age, infrequency of diagnoses such as psychotic disorders. sex, substance disorders, and psychopathology (major In conclusion, our results suggest that violent method of psychiatric axis I disorders as a categorical variable), we suicide could be a valid behavioral marker of lifetime im- found that history of lifetime aggression (p=0.03) and the pulsive aggression. This relationship is probably mediated interaction between lifetime aggression and lifetime im- by several other factors that this study did not aim to as- pulsivity (p<0.06) remained associated with a violent sess and that remain to be investigated further. Additional method of suicide. research including suicide attempters with a larger num-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1377 BRIEF REPORTS ber of female subjects who committed suicide and more meal? some thoughts on present research strategies. Pharma- individuals who committed suicide by nonviolent meth- copsychiatry 2002; 35:44–49 12. Lester D: Factors affecting choice of method of suicide. J Clin ods is needed to confirm and further understand the rela- Psychol 1970; 26:437 tionship between impulsive-aggressive behaviors, suicide 13. Kelly TM, Mann JJ: Validity of DSM-III-R diagnosis by psycholog- method, and suicide outcome. ical autopsy: a comparison with clinician ante-mortem diagnosis. Acta Psychiatr Scand 1996; 94:337–343 Received April 12, 2004; revised July 11 and Aug. 17, 2004; ac- 14. Conner KR, Conwell Y, Duberstein PR: The validity of proxy- cepted Sept. 9, 2004. From the McGill Group for Suicide Studies; and based data in suicide research: a study of patients 50 years of the Centre de Recherche Fernand-Seguin, Hôpital L-H Lafontaine, age and older who attempted suicide, II: life events, social sup- Université de Montréal, Montreal, Canada. Address correspondence port and suicidal behavior. Acta Psychiatr Scand 2001; 104: and reprint requests to Dr. Turecki, McGill Group for Suicide Studies, 452–457 Douglas Hospital, McGill University, 6875 LaSalle Blvd., Montreal QC 15. Zhang J, Conwell Y, Wieczorek WF, Jiang C, Jia S, Zhou L: Study- H4H 1R3 Canada; [email protected] (e-mail). ing Chinese suicide with proxy-based data: reliability and valid- Supported in part by Canadian Institutes of Health Research (CIHR) ity of the methodology and instruments in China. J Nerv Ment grant MOP-53321 and the Fonds de la Recherche Santé du Québec. Dr. Turecki is a CIHR scholar. Dis 2003; 191:450–457 16. Spitzer RL, Williams JBW, Gibbon M, First MB: The Structured Clinical Interview for DSM-III-R (SCID), I: history, rationale, and References description. Arch Gen Psychiatry 1992; 49:624–629 17. Chambers WJ, Puig-Antich J, Hirsch M, Paez P, Ambrosini PJ, 1. Suicide in the United States. Atlanta, Centers for Disease Con- Tabrazi MA, Davies M: The assessment of affective disorders in trol and Prevention, 2003 children and adolescents by semistructured interview: test-re- 2. Denning DG, Conwell Y, King D, Cox C: Method choice, intent, test reliability of the Schedule for Affective Disorders and and gender in completed suicide. Suicide Life Threat Behav Schizophrenia for School-Age Children, Present Episode Ver- 2000; 30:282–288 sion. Arch Gen Psychiatry 1985; 42:696–702 3. Murase S, Ochiai S, Ueyama M, Honjo S, Ohta T: Psychiatric fea- 18. Kovacs M: The Children’s Depression Inventory (CDI). Psycho- tures of seriously life-threatening suicide attempters: a clinical pharmacol Bull 1985; 21:995–998 study from a general hospital in Japan. J Psychosom Res 2003; 19. Kim C, Lesage A, Seguin M, Chawky N, Vanier C, Lipp O, Turecki 55:379–383 G: Patterns of co-morbidity in male suicide completers. Psy- 4. Self-directed violence, in World Report on Violence. Edited by chol Med 2003; 33:1299–1309 Krug GE, Dahlberg LL, Mercy JA, Zwi AB, Lozano R. Geneva, 20. Lesage AD, Boyer R, Grunberg F, Vanier C, Morissette R, Me- World Health Organization, 2002, pp 183–212 nard-Buteau C, Loyer M: Suicide and mental disorder: a case- 5. Heilä H, Isometsä ET, Henriksson MM, Heikkinen ME, Mart- control study of young men. Am J Psychiatry 1994; 151:1063– tunen MJ, Lönnqvist JK: Suicide and schizophrenia: a nation- 1068 wide psychological autopsy study on age- and sex-specific clin- 21. Brown GL, Goodwin FK, Ballenger JC, Goyer PF, Major LF: Ag- ical characteristics of 92 suicide victims with schizophrenia. gression in humans correlates with cerebrospinal fluid amine Am J Psychiatry 1997; 154:1235–1242 metabolites. Psychiatry Res 1979; 1:131–139 6. Maes M, Cosyns P, Meltzer HY, De Meyer F, Peeters D: Seasonal- 22. Barratt ES: Factor analysis of some psychometric measures of ity in violent suicide but not in nonviolent suicide or homicide. impulsiveness and anxiety. Psychol Rep 1965; 16:547–554 Am J Psychiatry 1993; 150:1380–1385 23. Cloninger CR, Przybeck TR, Svrakic DM, Wetzel RD: The Tem- 7. Conwell Y, Duberstein PR, Cox C, Herrmann J, Forbes N, Caine perament and Character Inventory (TCI): A Guide to Its Devel- ED: Age differences in behaviors leading to completed suicide. opment and Use. St Louis, Washington University, Center for Am J Geriatr Psychiatry 1998; 6:122–126 Psychobiology of Personality, 1994 8. Åsberg M, Traskman L, Thoren P: 5-HIAA in the cerebrospinal 24. Brent DA, Johnson BA, Perper J, Connolly J, Bridge J, Bartle S, fluid: a biochemical suicide predictor? Arch Gen Psychiatry Rather C: Personality disorder, personality traits, impulsive vio- 1976; 33:1193–1197 lence, and completed suicide in adolescents. J Am Acad Child 9. Brown GL, Ebert MH, Goyer PF, Jimerson DC, Klein WJ, Bunney Adolesc Psychiatry 1994; 33:1080–1086 WE, Goodwin FK: Aggression, suicide, and serotonin: relation- 25. Plutchik R: Outward and inward directed aggressiveness: the ships to CSF amine metabolites. Am J Psychiatry 1982; 139: interaction between violence and suicidality. Pharmacopsychi- 741–746 atry 1995; 28(suppl 2):47–57 10. Kullgren G, Tengstrom A, Grann M: Suicide among personality- 26. Shenassa ED, Catlin SN, Buka SL: Lethality of firearms relative disordered offenders: a follow-up study of 1943 male criminal to other suicide methods: a population based study. J Epide- offenders. Soc Psychiatry Psychiatr Epidemiol 1998; 33(suppl miol Community Health 2003; 57:120–124 1):S102–S106 27. Beautrais AL: Suicide and serious suicide attempts in youth: a 11. Muller-Oerlinghausen B, Roggenbach J: Concretism in biologi- multiple-group comparison study. Am J Psychiatry 2003; 160: cal suicide research—are we eating the menu instead of the 1093–1099

1378 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 Brief Report

Prediction of Panic Response to a Respiratory Stimulant by Reduced Orbitofrontal Cerebral Blood Flow in Panic Disorder

Justine M. Kent, M.D. pathophysiology of panic disorder. The aim of this study was to Jeremy D. Coplan, M.D. examine regional cerebral blood flow changes in response to anxiety/panic provocation in subjects with panic disorder and Osama Mawlawi, Ph.D. healthy comparison subjects. Jose M. Martinez, M.A. Method: Quantitative water method positron emission tomog- Susan T. Browne, B.A. raphy was used to obtain brain images of five untreated sub- Mark Slifstein, Ph.D. jects with panic disorder and five healthy comparison subjects Diana Martinez, M.D. before and during anxiogenic challenge with intravenous dox- Anissa Abi-Dargham, M.D. apram, an acute respiratory stimulant. Marc Laruelle, M.D. Results: Baseline perfusion of the orbitofrontal cortex pre- Jack M. Gorman, M.D. dicted panic attacks: lower perfusion was associated with heightened anxiety in response to doxapram challenge. Objective: Lack of appropriate top-down governance by fron- Conclusions: The orbitofrontal cortex may be important in the tal cortical regions over a hypersensitive amygdala-centered regulation of responding to fear and is a potential area of aber- fear neurocircuitry has been hypothesized to be central in the rant functioning in panic disorder.

(Am J Psychiatry 2005; 162:1379–1381)

A lthough much is now known about the pathophysi- in a psychiatric interview and the Structured Clinical Interview for ology of panic disorder, the neurocircuitry central to panic DSM-IV Axis I Disorders (6), and all were determined to be in good is still being delineated. The role of inhibitory cortical in- physical health after a complete medical evaluation. Subjects with panic disorder were included if they met DSM-IV criteria for panic puts on amygdalofugal pathways in the development of disorder; no other current axis I diagnoses were allowed except panic anxiety is especially poorly understood. dysthymia. Subjects with panic disorder were excluded if they had Although the results are not completely consistent, the a lifetime history of schizophrenia or bipolar disorder. Healthy majority of functional imaging studies in panic disorder comparison subjects did not meet criteria for any current or lifetime history of an axis I disorder. All subjects were required to be support frontal cortical deactivation during panic anxiety free of psychotropic medications for 6 weeks before scanning. The provoked by pharmacological challenges (1–3) and during study was approved by the institutional review boards of New York spontaneous panic attacks (4). Because the medial and or- State Psychiatric Institute and Columbia University. After com- bitofrontal regions of the prefrontal cortex are known to plete description of the study to the subjects, written informed have extensive connections with the amygdala, exerting a consent was obtained from all who participated. Subjects were scanned with [15O]H O PET on an ECAT EXACT primarily inhibitory influence on its activity (5), they are of 2 HR+ PET scanner (Siemens CTI, Knoxville, Tenn.); input function prime interest as areas of potentially aberrant functioning was measured with arterial blood samples. The respiratory stimu- in panic disorder. lant doxapram was used to induce anxiety in subjects (7). The The aims of this study were to use quantitative water pharmacological action of doxapram hydrochloride is mediated 15 through the peripheral carotid chemoreceptors, producing respi- method ([ O]H2O) positron emission tomography (PET) ratory stimulation within 20–40 seconds of injection, with a peak to test the following hypotheses: 1) Subjects with panic effect at 1–2 minutes and a total duration of action of approxi- disorder demonstrate greater activation of the amygdala mately 5 minutes. Because the effects of doxapram preclude a than healthy comparison subjects in response to an anxi- double-blind randomized study design, a single-blind nonran- ety-provoking challenge. 2) Subjects with panic disorder domized design was used. demonstrate less cortical restraint than healthy compari- All subjects underwent five activations (scans), each consisting of an intravenous bolus injection of 20 mCi [15O]H O. Scans were son subjects over subcortical fear pathway structures, as 2 acquired 15 minutes apart in a fixed order: resting baseline, re- manifest by hypoactivity in medial and orbitofrontal corti- peat resting baseline, placebo injection, placebo injection, cal regions. doxapram injection (0.5 mg/kg). Scanning began 30 seconds after placebo and doxapram injections. Throughout the scanning Method session, repeated anxiety assessments were made with three scales: 1) Acute Panic Inventory (8), 2) 10-point Anxiety Scale (8), Five subjects with a DSM-IV diagnosis of panic disorder (four and 3) 10-point Borg Breathlessness Scale (9). Occurrence of female, one male; mean age=27.2 years, SD=3.2) and five healthy panic was judged on the basis of DSM-IV criteria of a crescendo of comparison subjects (three female, two male; mean age=31.6 fear/anxiety with four or more associated symptoms (rated on the years, SD=6.8) participated in the study. All subjects participated Acute Panic Inventory).

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1379 BRIEF REPORTS

FIGURE 1. Relationship of Baseline Orbitofrontal Cerebral cebo or doxapram injections and no significant differ- Blood Flow (CBF) to Scores on the Acute Panic Inventory ences between the groups when regions of interest were During Doxapram Challenge in Four Patients With Panic Disorder and Five Healthy Comparison Subjects analyzed individually (Bonferroni correction for multiple comparisons). 50 When all subjects were considered together (N=9), baseline orbitofrontal CBF distinguished subjects who panicked (N=5) from those who did not panic (N=4) in 45 response to doxapram injection (t=2.82, df=7, p<0.05). Ad- ditionally, baseline orbitofrontal CBF was negatively correlated with anxiety scores on the Acute Panic Inventory 40 (r=–0.83, N=9, p<0.05) (Figure 1), the 10-point Anxiety Scale (r=–0.77, N=9, p<0.05), and the Borg Breathlessness < 35 Scale (r=–0.75, N=9, p 0.05) in response to doxapram injection.

30 Conclusions Orbitofrontal CBF (ml/min/100 g) Orbitofrontal Comparison subject Perfusion of the orbitofrontal prefrontal cortex at base- Panic disorder patient 25 line predicted panic vulnerability to respiratory challenge 0 10203040506070across subjects, suggesting that input from this cortical re- Acute Panic Inventory Score gion may be important in suppressing fear responding. In addition, orbitofrontal blood flow at baseline correlated negatively with scores on the Acute Panic Inventory, 10- Dynamic PET imaging data were summed and coregistered with magnetic resonance imaging (MRI). Data were fitted to the point Anxiety Scale, and Borg Breathlessness Scale in re- integrated form of the Kety-Schmidt equation (10) according to a sponse to doxapram challenge: higher levels of CBF in the lookup table. Summed PET data (from 30 seconds to 2 minutes af- orbitofrontal region was associated with lower anxiety ter injection) were matched to flow values from a lookup table scores and less breathlessness. generated by fitting the equations to a range of possible summed These data should be viewed cautiously given the small activity values. The delay between the time of measurement of arterial tracer concentration and the time of brain exposure was in- number of subjects and the high risk of a type II error. This corporated by fitting whole brain data to the dynamic form of the may account for our inability to demonstrate a relation- equations and including delay time as a fitted parameter. Para- ship between panic anxiety and amygdala activation. Lack metric maps were generated, with one flow value for each voxel. of differentiation between the panic and comparison Regions of interest were drawn on individual patients’ MRIs and groups in rCBF response to the anxiogenic challenge was transferred to the blood flow maps, generating regional mean perhaps due to a “floor” effect, in which the respiratory re- flow values. Subcortical regions of interest included the midbrain, hippocampus, amygdala, thalamus, and striatum. The prefrontal sponse to doxapram, resulting in hyperventilation, hypo- cortex was divided into the following regions of interest: anterior capnia, and resultant physiological vasoconstriction, cingulate, medial, dorsolateral, orbitofrontal, and subgenual. overshadowed potential differences between groups. This Change in cerebral blood flow (CBF) over time (scans 1–5) was potential confound of hypocapnia-induced vasoconstric- analyzed by using repeated-measures analysis of variance. Be- tion could be controlled for by measuring PCO2 with serial havioral data were compared with regional CBF (rCBF) values blood gases in future studies. Hypocapnia-induced vaso- from the scans by using correlational analyses based on the a priori hypotheses described above. constriction, however, is likely to be a global effect that would not account for the regional difference identified in Results the orbitofrontal cortex at baseline. Despite the limitations of this study, these results may stimulate more stud- All five of the subjects with panic disorder panicked in ies focusing on the orbitofrontal cortex and directed at de- response to doxapram, compared with one of the five termining the nature of inhibitory cortical input to the healthy subjects, a significant difference (χ2=5.76, df=1, amygdala in the development of panic vulnerability. p<0.02). One of the subjects with panic disorder was dropped from the PET data analysis because of a technical Received Oct. 23, 2003; revision received June 15, 2004; accepted problem with processing her PET data (this subject was 22 July 7, 2004. From the Department of Psychiatry and Department of Radiology, Columbia University College of Physicians & Surgeons; years old). There were no significant differences in rCBF in New York State Psychiatric Institute, New York; State University of the amygdala in response to placebo or doxapram injec- New York-Downstate, Brooklyn, N.Y.; and Mount Sinai School of Med- tions between the subjects with panic disorder and the icine, New York. Address correspondence and reprint requests to Dr. Kent, Department of Psychiatry, Unit 41, Columbia University, 1051 comparison subjects. Similarly, there were no significant Riverside Dr., New York, NY 10032; [email protected] (e-mail). differences between the groups in global (weighted aver- Supported by Eli Lilly & Co. age) subcortical or cortical blood flow in response to pla-

1380 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BRIEF REPORTS

References 5. Garcia R, Vouimba R-M, Baudry M, Thompson R: The amygdala modulates prefrontal cortex activity relative to conditioned 1. Woods SW, Koster K, Krystal JK, Smith EO, Zubal IG, Hoffer PB, fear. Nature 1999; 402:294–296 Charney DS: Yohimbine alters regional cerebral blood flow in 6. Spitzer RL, Williams JBW, Gibbon M, First MB: Structured Clini- panic disorder (letter). Lancet 1988; 2:678 cal Interview for DSM-IV (SCID). New York, New York State Psy- 2. Stewart RS, Devous MD Sr, Rush AJ, Lane L, Bonte FJ: Cerebral chiatric Institute, Biometrics Research, 1995 blood flow changes during sodium-lactate-induced panic at- 7. Lee Y, Curtis G, Weg J, Abelson J, Modell J, Campbell K: Panic attacks. Am J Psychiatry 1988; 145:442–449 tacks induced by doxapram. Biol Psychiatry 1993; 33:295–297 8. Dillon D, Gorman J, Liebowitz M, Fyer A, Klein D: Measurement 3. Boshuisen ML, Ter Horst GJ, Paans AM, Reinders AA, den Boer of lactate-induced panic and anxiety. Psychiatry Res 1987; 20: JA: rCBF differences between panic disorder patients and con- 97–105 trol subjects during anticipatory anxiety and rest. Biol Psychia- 9. Borg G: Psychophysical bases of perceived exertion. Med Sci try 2002; 52:126–135 Sports Exerc 1982; 14:377–381 4. Fischer H, Andersson J, Furmark T, Fredrikson M: Brain corre- 10. Kety SS, Schmidt CF: The determination of cerebral blood flow lates of an unexpected panic attack: a human positron emis- in man by the use of nitrous oxide in low concentrations. Am J sion tomographic study. Neurosci Lett 1998; 251:137–140 Physiol 1945; 143:53–66

Brief Report

Acute Stress Disorder and Posttraumatic Stress Disorder in Children and Adolescents Involved in Assaults or Motor Vehicle Accidents

Richard Meiser-Stedman, Ph.D. months, they reinterviewed 64 (68.8%) of the patients to assess William Yule, Ph.D. PTSD. Patrick Smith, Ph.D. Results: At initial interview, 18 (19.4%) of the 93 patients had Ed Glucksman, F.F.A.E.M. acute stress disorder and 23 (24.7%) met all acute stress disor- Tim Dalgleish, Ph.D. der criteria except dissociation. At 6 months, eight of the 64 patients (12.5%) had PTSD. Acute stress disorder and PTSD did not differ in prevalence between patients who had been assaulted Objective: The authors investigated acute stress disorder and later posttraumatic stress disorder (PTSD) in children and ado- and those who had been in accidents. Sensitivity and specificity lescents who had been involved in assaults or motor vehicle ac- statistics and regression modeling revealed that the diagnosis of cidents. acute stress disorder was a good predictor of later PTSD but that dissociation did not play a significant role. Method: They interviewed 93 patients 10–16 years old who were seen in an emergency department for having been as- Conclusions: Acute stress disorder has merit as a predictor of saulted or involved in a motor vehicle accident within 4 weeks later PTSD in children and adolescents, but dissociation has after the assault or accident to assess acute stress disorder. At 6 questionable utility.

(Am J Psychiatry 2005; 162:1381–1383)

Psychological trauma occurs at high rates in children at least three of a possible five dissociation symptoms. An and adolescents (1), and posttraumatic stress disorder important public health marker of the utility of acute (PTSD) in this age group has attracted considerable clini- stress disorder is its ability to predict later PTSD, thus al- cal and research interest. However, the diagnosis of acute lowing clinicians to focus resources on susceptible indi- stress disorder, introduced in DSM-IV, has received rela- viduals (4). In adults, acute stress disorder is a good pre- tively little attention in younger populations (2), despite a dictor of later PTSD (5, 6), but the dissociation symptoms growing body of research in adults (3). Unlike PTSD, which appear to add little (7). is diagnosed at least 4 weeks after trauma, acute stress dis- Several studies have examined acute stress disorder order is diagnosed 2 days to 4 weeks after trauma. Acute symptoms in younger populations (8–12). However, only stress disorder also differs from PTSD in being explicitly one study (11) examined the power of the acute stress dis- conceived as a dissociative response to trauma requiring order diagnosis (derived solely from questionnaire re-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1381 BRIEF REPORTS

Brief Report

Acute Stress Disorder and Posttraumatic Stress Disorder in Children and Adolescents Involved in Assaults or Motor Vehicle Accidents

(Am J Psychiatry 2005; 162:1381–1383)

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1381 BRIEF REPORTS

TABLE 1. Sensitivity, Specificity, and Positive and Negative Predictive Power of Acute Stress Disorder, Subacute Stress Disorder, and Early PTSD Criteria and Diagnoses to Predict Later PTSD Among Children and Adolescents Involved in Assaults or Motor Vehicle Accidentsa

Frequency Positive Negative Cases Correctly Criterion/Diagnosis N%Sensitivity Specificity Predictive Power Predictive Power Allocated (%) Acute stress disorder criteria A. Traumatic stressor 67 72.0 1.00 0.32 0.17 1.00 40.6 B. Dissociation (at least three symptoms) 50 53.8 0.75 0.55 0.19 0.94 57.8 C. Reexperiencing (at least one symptom) 63 67.7 1.00 0.30 0.17 1.00 39.1 D. Avoidance (at least one symptom) 63 67.7 0.88 0.34 0.16 0.95 40.6 E. Arousal (at least one symptom) 75 80.6 1.00 0.23 0.16 1.00 32.8 F. Impairment 36 38.7 0.75 0.64 0.23 0.95 65.6 Acute stress disorder diagnosis (A + B + C + D + E + F) 18 19.4 0.50 0.88 0.36 0.92 82.8 Subacute stress disorder diagnosis (A + C + D + E + F) 23 24.7 0.63 0.82 0.33 0.94 79.7 Early PTSD criteria Reexperiencing (at least one symptom) 69 74.2 1.00 0.25 0.16 1.00 34.3 Avoidance (at least three symptoms) 49 52.7 0.50 0.45 0.11 0.86 45.3 Arousal (at least two symptoms) 60 64.5 0.88 0.38 0.17 0.95 43.8 Early PTSD diagnosis (early PTSD criteria + A + F) 23 24.7 0.63 0.79 0.29 0.94 76.6 PTSD diagnosis at 6-month assessment (N=64) 8 12.5 a Sensitivity=the probability that someone diagnosed with PTSD at follow-up had a given diagnosis, etc., at initial interview; Specificity=the probability that someone not diagnosed with PTSD at follow-up did not have a given diagnosis, etc., at initial interview; positive predictive power=the probability that someone who has a given diagnosis, etc., goes on to have a diagnosis of PTSD at 6-month follow-up; and negative predictive power=the probability that someone who does not have a given diagnosis, etc., does not go on to have a diagnosis of PTSD at 6- month follow-up. sponses) to predict later PTSD. Results indicated that pants reinterviewed at 6 months were no more or less likely than acute stress disorder had low predictive sensitivity (unlike participants not reinterviewed to meet criteria for initial acute in adults) but that this was improved by removal of the dis- stress disorder. After description of the study, written informed consent from both children and their caregivers was obtained for sociation items. Further research on the predictive utility all participants. of acute stress disorder in younger samples is urgently The child version of the Anxiety Disorders Interview Schedule needed. for DSM-IV (13) was used to diagnose acute stress disorder and To our knowledge, the present study is the first to assess PTSD. This instrument is a structured interview schedule with acute stress disorder in children and adolescents using good psychometric properties (13, 14) for the assessment of anxiety disorders in young subjects. It does not include the dissocia- structured clinical interview. Participants were assessed 2– tive symptoms of acute stress disorder. Therefore, we designed a 4 weeks and 6 months after physical assault or motor vehi- number of supplementary interview items (available from R.M.-S.) cle accident. These events were selected because both are to assess these symptoms, guided by DSM-IV and existing adult common, single-incident traumas with comparable rates acute stress disorder interview schedules (15). of acute stress disorder in adults (5, 6). Our goal was to Subacute stress disorder was defined as the full acute stress disorder diagnosis minus the dissociation criterion. Early PTSD was compare the utility of diagnoses of acute stress disorder, defined as PTSD at initial assessment minus the duration crite- subacute stress disorder (acute stress disorder minus the rion (7). Internal reliabilities for acute stress disorder, subacute dissociation criterion) (3), and “early PTSD” (PTSD with- stress disorder, and early PTSD diagnoses in the current study out the duration criterion) (7) in predicting later PTSD. were high (Cronbach’s alpha=0.85–0.87). There was unanimous diagnostic agreement between independent raters for 11 initial Method and 10 follow-up interviews (kappa=1.00). Children and adolescents (10–16 years old) who were treated in Results a London emergency department following motor vehicle accident or assault met study criteria (N=343). One hundred nineteen Table 1 shows how many children and adolescents met (34.7%) of these patients could not be contacted because of in- criteria for acute stress disorder, subacute stress disorder, complete emergency department records, two (0.6%) were imme- and early and later PTSD. Chi-square analyses revealed no diately referred for treatment, and 116 (33.8%) declined to partic- significant differences in prevalence for any diagnosis be- ipate. The 106 children and adolescents (30.9%) who consented tween those exposed to assaults and those involved in mo- to participate were assessed within 4 weeks. Of these, 93 (87.7%) completed an initial clinical interview and 64 (60.4%) completed tor vehicle accidents. a second interview at 6 months. The mean age of the 93 patients Table 1 also shows the sensitivity, specificity, and posi- who completed the initial interview was 13.9 years (SD=1.9); 33 tive and negative predictive power of the individual symp- were girls. tom criteria and the acute stress disorder, subacute stress There was no difference between participants and nonpartici- disorder, and early PTSD diagnoses at initial interview to pants in terms of sex, type of trauma, injury severity, or ethnicity, although participants were significantly younger (nonpartici- predict later PTSD, as well as the number of later PTSD pants’ mean age=14.8, SD=1.9) (t=1.98, df=364.1, p<0.05). Partici- cases that were correctly diagnosed by each criterion or

1382 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BRIEF REPORTS diagnosis. Subacute stress disorder was the diagnosis that The authors thank Nicola Leete and other staff at King’s College gave the best balance of sensitivity and specificity. Hospital, London, for support in recruiting participants and Dr. Sean Perrin for training given to Dr. Meiser-Stedman. We used logistic regression to examine whether the acute stress disorder dissociation criterion explained any unique variance in later PTSD. Subacute stress disorder References was entered in the first step, resulting in a significant model (χ2=6.56, df=1, p<0.01) and accounting for unique 1. Giaconia RM, Reinherz HZ, Silverman AB, Pakiz B, Frost AK, Co- hen E: Traumas and posttraumatic stress disorder in a commu- variance (Wald χ2=6.33, df=1, p<0.01). Entering the disso- nity population of older adolescents. J Am Acad Child Adolesc ciation criterion in the second step did not significantly Psychiatry 1995; 34:1369–1380 improve the model’s ability to predict later PTSD and did 2. March JS: Acute stress disorder in youth: a multivariate predic- not account for any unique variance. tion model. Biol Psychiatry 2003; 53:809–816 3. Harvey AG, Bryant RA: Acute stress disorder: a synthesis and Discussion critique. Psychol Bull 2002; 128:886–902 4. Marshall RD, Spitzer R, Liebowitz MR: Review and critique of The rate of acute stress disorder in this group of children the new DSM-IV diagnosis of acute stress disorder. Am J Psychi- and adolescents (19.4%) is similar to that found in adults atry 1999; 156:1677–1685 (5, 6) but slightly higher than in other child studies (8, 11). 5. Brewin CR, Andrews B, Rose S, Kirk M: Acute stress disorder As in adult studies, acute stress disorder occurred at simi- and posttraumatic stress disorder in victims of violent crime. lar rates among subjects involved in assaults or motor ve- Am J Psychiatry 1999; 156:360–366 hicle accidents (5, 6). 6. Harvey AG, Bryant RA: The relationship between acute stress The acute stress disorder diagnosis was a good predictor disorder and posttraumatic stress disorder: a prospective eval- of later PTSD at follow-up, correctly classifying 82.8% of uation of motor vehicle accident survivors. J Consult Clin Psy- chol 1998; 66:507–512 PTSD cases. However, subacute stress disorder provided a 7. Brewin CR, Andrews B, Rose S: Diagnostic overlap between better balance between sensitivity and specificity at pre- acute stress disorder and PTSD in victims of violent crime. Am dicting later PTSD than full acute stress disorder. Further- J Psychiatry 2003; 160:783–785 more, regression analysis revealed that the dissociation 8. Bryant B, Mayou R, Wiggs L, Ehlers A, Stores G: Psychological criterion did not significantly enhance the ability of sub- consequences of road traffic accidents for children and their acute stress disorder to predict later PTSD. As in the adult mothers. Psychol Med 2004; 34:335–346 literature (7), acute stress disorder and early PTSD were 9. Daviss WB, Racusin R, Fleischer A, Mooney D, Ford JD, McHugo equally effective predictors of later PTSD. GJ: Acute stress disorder symptomatology during hospitalization for pediatric injury. J Am Acad Child Adolesc Psychiatry The study had two important limitations. First, the rela- 2000; 39:569–575 tively small number of subjects may have weakened sta- 10. Fein JA, Kassam-Adams N, Vu T, Datner EM: Emergency depart- tistical power. Second, the study group comprised older ment evaluation of acute stress disorder symptoms in violently children (essentially preadolescents) and adolescents. injured youths. Ann Emerg Med 2001; 38:391–396 Studies examining acute stress disorder in larger and 11. Kassam-Adams N, Winston FK: Predicting child PTSD: the rela- younger populations are therefore needed. tionship between acute stress disorder and PTSD in injured In conclusion, the acute stress disorder diagnosis is a children. J Am Acad Child Adolesc Psychiatry 2004; 43:403–411 good predictor of later PTSD in children and adolescents. 12. Winston FK, Kassam-Adams N, Vivarelli-O’Neill C, Ford JD, New- However, our data indicate no unique role for the dissoci- man E, Baxt C, Stafford P, Cnaan A: Acute stress disorder symptoms in children and their parents after pediatric traffic injury. ation criterion of acute stress disorder in these patients. Pediatrics 2002; 109:e90 13. Silverman WK, Albano AM: Anxiety Disorders Interview Sched- Received July 30, 2003; revised Feb. 10 and July 14, 2004; accepted ule for DSM-IV: Child and Parent Interview Schedule. San Anto- Aug. 30, 2004. From the Department of Psychology, Institute of Psychiatry. Address correspondence and reprint requests to Dr. nio, Tex, Psychological Corp, 1996 Meiser-Stedman, Department of Psychology, Institute of Psychiatry, 14. Silverman WK, Saavedra LM, Pina AA: Test-retest reliability of De Crespigny Park, London, SE5 8AF, U.K.; r.meiser-stedman@ anxiety symptoms and diagnoses with the Anxiety Disorders iop.kcl.ac.uk (e-mail). Interview Schedule for DSM-IV: child and parent versions. J Am Supported in part by a Medical Research Council of the United Acad Child Adolesc Psychiatry 2001; 40:937–944 Kingdom (UKMRC) Research Studentship (G78/6730) and a Peggy Pol- lak Fellowship from the Psychiatry Research Trust (Dr. Meiser-Sted- 15. Bryant RA, Harvey AG: Acute Stress Disorder: A Handbook of man); the UKMRC (Dr. Dalgleish); and a Kraupl Taylor Fellowship Theory, Assessment, and Treatment. Washington, DC, Ameri- from the Psychiatry Research Trust (Dr. Smith). can Psychological Association, 2000

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1383 Letters to the Editor

48,XXYY Syndrome, Mood Disorder, 5. Maclean N, Mitchell JM: A survey of sex-chromosome abnor- and Aggression malities among 4,514 mental defectives. Lancet 1962; 1:293– 296 TO THE EDITOR: A man with 48,XXYY syndrome was evaluated 6. Sorensen K, Nielsen J, Jacobsen P, Rolle T: The 48,XXYY syn- for suicidal and aggressive behavior. This unusual syndrome drome. Ment Defic Res 1978; 22:197–205 alerted our service to the associated psychopathology in this 7. Zelante L, Piemontese MR, Francioli G, Clavano S: Two 48,XXYY genetic disorder. patients: clinical, cytogenetic and molecular aspects. Ann Genet 2003; 46:479–481 Mr. A, a 24-year-old man with a documented chromo- SERMSAK LOLAK, M.D. somal abnormality of 48,XXYY evaluated by a standard cy- ELISA DANNEMILLER, B.S. togenetics technique, was admitted to the hospital for sui- FRANCIS ANDRES, M.D. cidal ideation and aggressive behavior toward his brother, Falls Church, Va. whom he pushed down the stairs. He endorsed symptoms suggestive of a mixed bipolar episode before admission. He appeared to have borderline intelligence, with a docu- Dysbetalipoproteinemia and Clomipramine mented IQ of 70–80, and was able to finish high school. TO THE EDITOR: Approximately 1% of North Americans and There was no family history of psychiatric or genetic disorders. His physical appearance was notable for a tall stat- Northern Europeans are homozygous for apolipoprotein-E2 ure, gynecomastia, and truncal obesity. His mental status (APOE-2) (1). However, overt hyperlipoproteinemia occurs in examination was significant for irritability and a labile, in- only 2%–10% of these people, resulting in a markedly intense affect. His cognition was intact. He had minimal increased risk of atherosclerosis (1). There are numerous envi- sight into his illness. ronmental factors known to precipitate hyperlipoproteine- Upon admission, Mr. A was given olanzapine and was mia in APOE-2 homozygotes. These include the coexistence later switched to oxcarbazepine and citalopram. Provi- of other genetic dyslipidemias, hypothyroidism, diabetes sional diagnoses of bipolar disorder, not otherwise speci- mellitus, menopause, alcohol, poor diet, and obesity (1). Al- fied, and cluster B personality traits were made. Mr. A was though protease inhibitors have been observed to precipitate initially uncooperative and frequently became agitated. dysbetalipoproteinemia (2), there are no such reports impli- Toward the end of admission, Mr. A seemed to respond to cating tricyclic antidepressants. However, mild (3, 4) to mod- oxcarbazepine. His mood and behavior improved. He was discharged after 10 days of hospitalization with oxcarba- erate (5) increases induced by tricyclic antidepressants in to- zepine, 1200 mg/day, and citalopram, 40 mg/day. tal cholesterol have been observed in the general population with these medications. 48,XXYY syndrome was initially considered a variant of Klinefelter’s syndrome (1). Nowadays, it is accepted as a dis- Mr. A, a 48-year-old man with major depressive disor- tinct clinical and genetic entity (2, 3). Individuals with this der, was referred to our clinic for severe dyslipidemia. At the time, he was not taking lipid-lowering medications. syndrome are more aggressive, more intellectually handi- For most of the previous 20 years, he had been taking the capped, and taller than people with Klinefelter’s syndrome tricyclic antidepressant clomipramine, first at 75 mg/day (4). Children and adolescents with 48,XXYY syndrome often and then at 150 mg/day. Attempts to control his symp- come to psychiatrists for behavioral problems, sometimes toms with selective serotonin reuptake inhibitors had even before the chromosomal diagnosis is made. The inci- failed. His other medications were carbamazepine, 200 dences among newborn and institutionalized mentally re- mg t.i.d., and clonazepam, 1.5 mg at bedtime. tarded patients are 1/50,000 and 0.08%–0.33%, respectively Mr. A’s family physician had described his initial lipid el- (5, 6). Behavioral and intellectual disturbances in these pa- evations as “moderate” and discovered them about 5 tients are likely related to sex-chromosomal abnormalities years after the initiation of clomipramine. However, the (3). Clinical signs of 48,XXYY are generally nonspecific. Some increase to 150 mg/day resulted in severe dyslipidemia: a total cholesterol level of 694.9 mg/dl, a triglyceride level of the common features include tall stature, gynecomastia, of 637.2 mg/dl, a high-density lipoprotein cholesterol truncal obesity, skin ulcers, and a craniofacial dysmorphism level of 81.1 mg/dl, and a total cholesterol/high-density li- described as a “pugilistic” facial appearance (7). poprotein cholesterol ratio of 8.6. Of interest, before his To our knowledge, there are no reports of using psychotro- referral, Mr. A had implicated clomipramine by discon- pics specifically to treat behavioral problems in this patient tinuing it under his psychiatrist’s supervision. Dramatic population. improvements were observed after 11 days. His total cholesterol level was 338.2 mg/dl, his triglyceride level was References 210.6 mg/dl, his high-density lipoprotein cholesterol level 1. Muldal S, Ockey CH: The “double male”: a new chromosome was 51.0 mg/dl, his low-density lipoprotein cholesterol constitution in Klinefelter’s syndrome. Lancet 1960; 2:492–493 level was 245.6 mg/dl, and his total cholesterol/high-den- 2. Demirhan O: Clinical findings and phenotype in a toddler with sity lipoprotein cholesterol ratio was 6.6. 48,XXYY syndrome (letter). Am J Med Genet A 2003; 119:393– Within weeks, Mr. A’s depressive symptoms returned, 394 and clomipramine was reinitiated, again causing severe 3. Cammarata M, Di Simone P, Graziano L, Giuffre M, Corsello G, dyslipidemia. The lipid-raising effect of clomipramine was Garofalo G: Rare sex chromosome aneuploidies in humans: observed both with and without co-administration of report of six patients with 48,XXYY, 49,XXXXY, and 48,XXXX carbamazepine, implicating the tricyclic antidepressant karyotypes. Am J Med Genet 1999; 85:86–87 alone. Mr. A’s levels of thyroid-stimulating hormone and 4. Parker CE, Mavalwala J, Melnyk J, Fish CH: The 48,XXYY syn- fasting glucose were normal. His medical history was neg- drome. Am J Med 1970; 48:777–781 ative for diabetes mellitus, renal, and hepatic disease, ex-

1384 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 LETTERS TO THE EDITOR

cessive alcohol intake, and obesity. Although his father Ms. A was a 47-year-old white woman who was hospital- had had minor increases in total cholesterol, there was no ized for an episode of severe depression in the context of family history of coronary heart disease, peripheral or vas- a 20-year history of schizoaffective disorder. Ms. A had cular disease, or stroke among first-degree relatives. been treated for depression with ECT 8 years previously Mr. A’s physical examination was unremarkable except and had experienced profound confusion and short-term for the presence of palmar xanthomas, pathognomonic of memory problems. However, she did respond, with remis- dysbetalipoproteinemia (type III hyperlipoproteinemia). sion of her depression. The current episode of depression He had first noticed the orange discoloration of his palmar had persisted for 16 weeks, during which she did not re- creases 4 years after starting clomipramine but was un- spond to venlafaxine, bupropion, sertraline, olanzapine, aware of its significance. Genetic testing confirmed APOE- or lamotrigine. She had been simultaneously taking val- 2 homozygosity. Since lovastatin had previously failed to proic acid. Ms. A’s depression was characterized by pro- improve his lipid levels, fenofibrate, 160 mg/day, was pre- found dysphoria, anergia, anorexia, insomnia, anhedonia, scribed. Unfortunately, his mood soon began to deterio- and passive suicidal ideation. There was no observed or rate, and he had increased frequency of suicidal ideation. reported mood variability, except for mild morning wors- Thus, although it was unclear that fenofibrate was the ening of the dysphoria. She was hospitalized because of a cause, we elected to discontinue it. Atorvastatin, 40 mg/ decline in her activities of daily living to the point of not day, was then prescribed. Subsequently, his lipid levels dressing herself without assistance. were somewhat improved: his total cholesterol level was Ms. A complained of severe memory problems, al- 517.4 mg/dl, his triglyceride level was 460.2 mg/dl, his though her score on the Mini-Mental Status Examination high-density lipoprotein cholesterol level was 57.9 mg/dl, was 30 of 30. The results of an EEG and magnetic reso- and his total cholesterol/high-density lipoprotein choles- nance imaging were normal. Ms. A was withdrawn from terol ratio was 8.9. Of interest, Mr. A complained of a de- valproic acid, 1500 mg/day, and lamotrigine, 50 mg/day, pressed mood while taking atorvastatin, also necessitating 24 hours before her first ECT treatment. For the index ECT its discontinuation. He recently started taking ezetimibe treatment, ketamine was used as an induction agent at a for lowering his lipid levels. dose of 0.5 mg/kg because we have found that it reduces cognitive side effects in some patients. Ms. A had no pre- We suggest that the use of tricyclic antidepressants may in- vious exposure to ketamine. Bifrontal lead placement was duce dysbetalipoproteinemia in APOE-2 homozygotes. Ap- used, and a stimulus with the Spectrum 5000 Q ECT appa- propriate referral and APOE genotyping should be considered ratus (MECTA Corp., Lake Oswego, Ore.) was administered for patients discovered to be dyslipidemic while taking tri- by using the dose-titration method. The cuff method was cyclic antidepressants. used to monitor the motor seizure, and the electrical seizure was monitored with an EEG. No motor or electrical References seizure was observed during the first treatment session. 1. Mahley RW, Rall SC Jr: Type III hyperlipoproteinemia (dysbetali- Ms. A reported an immediate improvement of her poproteinemia): the role of apolipoprotein E in normal and mood after regaining consciousness. This improvement abnormal metabolism, in The Metabolic and Molecular Bases continued the next day when she awoke in the morning of Inherited Disease, 8th ed. Edited by Scriver CR, Beaudet AR, and reported a subjective improvement in well-being and Sly WS, Valle D. New York, McGraw-Hill, 2000, pp 2835–2862 an appetite for the first time in 2 weeks. The following 2. Lister RK, Youle M, Nair DR, Winder AF, Rustin MH: Latent dys- day, ECT was again administered with the dose-titration betalipoproteinaemia precipitated by HIV-protease inhibitors method because we assumed that the antiepileptic agents (letter). Lancet 1999; 353:1678 interfered with the induction of a seizure during the first 3. Yeragani VK, Balon R, Pohl R, Ramesh C: Imipramine treat- treatment session. Again, no electrical or motor seizure ment and increased serum cholesterol levels (letter). Can J Psy- was observed. Strikingly, Ms. A reported a further im- chiatry 1989; 34:845 provement in her mood that persisted the following day. 4. Yeragani VK, Pohl R, Balon R, Ramesh C, Glitz D: Increased se- Her core symptoms, interest, energy, motivation, mood, rum total cholesterol to HDL-cholesterol ratio after imipra- and appetite all improved. On our 0–10-point rating scale mine. Psychiatry Res 1990; 32:207–209 of clinical improvement, Ms. A rated herself at 7, having 5. Roessner V, Demling J, Bleich S: Doxepin increases serum cho- initially rated herself at 2. Session 2 fell on a Friday, so lesterol levels. Can J Psychiatry 2004; 49:74–75 there was an extra day to observe Ms. A’s response. DANIEL T. HOLMES, M.D. Forty-eight hours after her second ECT session, Ms. A PHILLIP LONG, M.D. still noted improvement. She did note some decline in her JIRI FROHLICH, M.D. mood and felt that the improvement was starting to dissi- Vancouver, B.C., Canada pate. Because of the timing of the treatments, we had 5 days to observe her clear improvement over baseline in response to ketamine. At session 3, a full grand mal sei- Antidepressant Effect of Ketamine During ECT zure was observed. Three more treatment sessions were necessary before remission was reached. TO THE EDITOR: There has been recent interest in the use of ketamine as anesthesia for ECT because of its low anticonvul- Ketamine, as both an adjunctive dose (3) and an inductive sant effects and possible reduction of cognitive side effects dose (4), has been noted to improve mood in patients under- (1). Ketamine, a noncompetitive N-methyl-D-aspartate going surgery. Two studies have demonstrated its potential (NMDA) glutamate receptor antagonist, has also been shown antidepressant effects in major depression (2, 3). Of interest, to have putative antidepressant effects (2–4). The following acute administration of NMDA antagonists has been demon- case serves to highlight the possible antidepressant effects of strated to induce neurogenesis, a characteristic seen in many ketamine in a patient who unintentionally received an induc- antidepressant agents. NMDA-mediated involvement of the tion dose alone during two failed ECT sessions. glutamatergic system in the pathophysiology of depression is

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1385 LETTERS TO THE EDITOR of growing interest (5, 6). Ketamine is also a weak dopamine weeks, Ms. A’s depression had responded to treatment, transporter antagonist and can be psychotomimetic. Its role with more than a 50% reduction of her score on the Hamil- as a primary anesthetic agent in ECT deserves more study. It ton Depression Rating Scale, while her WBC count was × 9 may have its greatest use in patients with severe nonpsychotic 6.1 10 /liter. Upon her final assessment, after 6 months of taking sertraline, her depression had remitted completely, depression. without any adverse effects. References Since there was neither concomitant medication nor med- 1. Krystal AD, Weiner RD, Dean MD, Lindahl VH, Tramontozzi LA, ical illness, an association between mirtazapine and severe Falcone G, Coffey CE: Comparison of seizure duration, ictal EEG, and cognitive effects of ketamine and methohexital anes- neutropenia might be suggested. Neutropenia with cross- thesia with ECT. J Neuropsychiatry Clin Neurosci 2003; 15:1:27– intolerance between two tricyclics has been described before 34 (6), and there is evidence that patients may successfully be 2. Berman RM, Cappiello A, Anand A, Oren DA, Henninger GR, treated with another class of drug after such an incidence (4). Charney DS, Krystal JH: Antidepressant effects of ketamine in Therefore, a selective serotonin reuptake inhibitor, sertraline, depressed patients. Biol Psychiatry 2000; 47:351–354 was administered and successfully used without an adverse 3. Kudoh A, Takahira Y, Katagai H, Takazawa T: Small-dose keta- effect. We may hypothesize that a different class of antide- mine improves the postoperative state of depressed patients. pressants might cause agranulocytosis by different mecha- Anesth Analg 2002; 95:114–118 nisms. Patients should be monitored closely for symptoms in- 4. Yilmaz A, Schulz D, Aksoy A, Canbeyli R: Prolonged effect of an dicating agranulocytosis. An antidepressant from a different anesthetic dose of ketamine on behavioral despair. Pharmacol Biochem Behav 2002; 71:349–352 class might be considered after such an incident. 5. Sanacora G, Rothman DI, Mason G, Krystal JH: Clinical studies References implementing glutamate neurotransmission in mood disor- 1. Albertini RS, Penders TM: Agranulocytosis associated with tricy- ders. Ann NY Acad Sci 2003; 1003:292–308 clics. J Clin Psychiatry 1978; 39:483–485 6. Paul IA, Skolnick P: Glutamate and depression: clinical and 2. Adams PC, Robinson A, Reid MM, Vishu MC, Livingston M: preclinical studies. Ann NY Acad Sci 2003; 1003:250–272 Blood dyscrasias and mianserin. Postgrad Med J 1983; 59:31– ROBERT OSTROFF, M.D. 33 MARBIELA GONZALES, M.D. 3. Nelson JC: Safety and tolerability of the new antidepressants. J GERARD SANACORA, M.D. Clin Psychiatry 1997; 58:26–31 Hamden, Conn. 4. Anghelescu I, Klawe C, Dahmen N: Venlafaxine in a patient with idiopathic leukopenia and mirtazapine-induced severe Sertraline- and Mirtazapine-Induced Severe neutropenia (letter). J Clin Psychiatry 2002; 63:838 Neutropenia 5. Davis J, Barkin RL: Clinical pharmacology of mirtazapine: revis- ited (letter). Am Fam Physician 1999; 60:1101 TO THE EDITOR: Neutropenia is a rare and reversible side effect 6. Draper BM, Manoharan A: Neutropenia with cross-intolerance of antidepressant treatment (1–4). Six cases of agranulocytosis between two tricyclic antidepressant agents (letter). Med J Aust in approximately 2 million exposures to mirtazapine have been 1987; 146:452–453 reported (unpublished data from Organon, Inc., 1999). Be- TUBA OZCANLI, M.D. cause all of the patients had either a concomitant medication BARIS UNSALVER, M.D. or disease that might have been related to agranulocytosis, it SAMURAY OZDEMIR, M.D. has been suggested that the association between mirtazapine MINE OZMEN, M.D. Istanbul, Turkey and agranulocytosis might have been coincidental (5). We report on a person with depression who developed severe neutropenia during treatment with mirtazapine and was Clozapine-Induced Eosinophilic Colitis safely treated with sertraline. TO THE EDITOR: Clozapine is an atypical antipsychotic with Ms. A, a 44-year-old woman with complaints of sleep considerable efficacy compared to other antipsychotic medi- disturbance, lack of energy, and unhappiness, was diag- cations (1). We report on a patient who developed fever and nosed with major depressive disorder and administered diarrhea while taking clozapine and was diagnosed with mirtazapine, 30 mg/day. Her medical history was nega- clozapine-induced eosinophilic colitis. tive, and the results of routine blood tests (WBC count of 6.8×109/liter) were unremarkable except for a total cho- Mr. A was a 45-year-old man with schizophrenia who lesterol level of 204 mg/dl. Three weeks later, she came in had psychotic decompensation in the setting of medica- with complaints of a severe sore throat, difficulty swallow- tion noncompliance and developed neuroleptic malig- ing, a loss of appetite, and an aphthous ulcer in her oral nant syndrome when haloperidol and risperidone were mucosa, with a slightly elevated body temperature of restarted. He was treated with ECT. Concurrently, cloza- 37.7°C during her physical examination. In subsequent pine was started at a low dose and gradually increased. blood counts, neutropenia (a WBC count of 2.2×109/liter) On the 14th day of clozapine therapy, he developed a fe- was detected, and a blood smear revealed a granulocyte ver of 103.6°F and profuse nonbloody diarrhea. His cloza- number of 1.1×109/liter. Mirtazapine was immediately dis- pine dose was 200 mg/day. His other medications at the continued, and sultamicillin, 375 mg b.i.d., was started af- time were lorazepam, aspirin, and metoprolol. His other ter consulting with the otorhinolaryngology department. vital signs were stable, and there was no muscle rigidity or Within 2 weeks, Ms. A’s WBC count and granulocyte count elevation of creatine kinase to suggest recurrence of neu- had gradually increased to 3.8×109/liter and 3.2×109/liter, roleptic malignant syndrome. respectively. Four weeks after discontinuation of mirtaza- His laboratory values showed mild elevation of his WBC pine, sertraline was administered at 50 mg/day. Within 6 count (to 12,300/mm3). The results of multiple blood cul-

1386 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 LETTERS TO THE EDITOR

tures, urine cultures, a chest X-ray, stool studies, and an to compare the efficacy of a low-dose antipsychotic with pla- HIV test were negative. Mr. A continued to have fever and cebo in preventing or delaying the onset of psychosis (2). profuse diarrhea in the ensuing 10 days. Laboratory stud- We present recruitment and decision data from the Calgary ies were significant for peripheral eosinophilia, with a site. During a 24-month recruitment period, 95 individuals peak of 22% and an absolute count of 2,140/mm3, as well as an elevated erythrocyte sedimentation rate of 116 mm/ were identified as potentially eligible for the study. Based on hour. A colonoscopy was performed on the 10th day of study entry criteria, 38% (N=36) were eligible; 86% of the eligi- the fever and diarrhea. Mr. A’s colon was normal upon ble persons refused to enter the trial. Of those eligible, 39% gross examination, and random biopsies were taken. Mi- who were troubled by their symptoms wanted active treat- croscopically, patchy eosinophilic infiltrate and histiocytic ment as soon as possible and were offered treatment in our aggregates, focally associated with crypt destruction, were Early Psychosis Program; 47% of the eligible subjects were noted in the architecturally preserved colonic mucosa. concerned, often felt debilitated by their symptoms, but did Clozapine was thought to be the culprit and discontin- not want to take medication. This group either consented to ued. Mr. A’s fever and diarrhea disappeared the next day; enter a 1-year observational study offering education, sup- his eosinophilia and erythrocyte sedimentation rate report, and monthly assessments or attended a session to returned to normal within several days. ceive education about their symptoms. All had immediate There have been reports of fever and diarrhea associated access to a clinician should they experience an increase in with clozapine use (2). However, the etiology of these symp- symptoms in the near future. The remaining 14% consented to toms has not been clear. Eosinophilic colitis has been sug- enter the medication trial. An assessment of presenting symp- gested, although there has never been any pathologic confir- toms with the Scale of Prodromal Symptoms (2) demonstrated mation of these findings (3). To our knowledge, this is the first that those who wanted treatment and those who entered the report of a comprehensive investigation of the etiology with trial had significantly higher scores on both the positive and extensive laboratory studies as well as microscopic visualiza- negative symptom scales of the Scale of Prodromal Symptoms tion of colonic pathology. We note the presence of an elevated (p<0.05) than those who did not want active treatment. erythrocyte sedimentation rate in this case, as is often seen in Clinical trials examining the impact of antipsychotic medi- eosinophilic colitis (4). Histological examination of eosino- cation on preventing or delaying schizophrenia psychoses philic colitis usually shows patchy clusters or sheets of likely recruit a small proportion of eligible individuals. The eosinophils in the lamina propria and crypt epithelium, with fact that those who enter such a trial have greater symptoms or without crypt destruction (5). The histopathology observed than those who refused to participate in the trial demon- in our patient was consistent with eosinophilic colitis. strates that putatively prodromal individuals who enter med- References ication trials are sick individuals with significant impairment who may potentially be on the cusp of conversion. The major- 1. Baldessarini RJ, Frankenburg FR: Clozapine, a novel antipsy- ity of eligible subjects who have less marked symptoms prefer chotic agent. N Engl J Med 1991; 324:746–754 to be involved to some degree in a range of educational and 2. Patterson BD, Jennings JL: Spiking fever and profuse diarrhea with clozapine treatment (letter). Am J Psychiatry 1993; 150: psychological interventions. 1126 References 3. Friedberg JW, Frankenburg FR, Burk J, Johnson W: Clozapine- caused eosinophilic colitis. Ann Clin Psychiatry 1995; 7:97–98 1. McGorry PD, Yung AR, Phillips LJ, Yuen HP, Francey S, Cosgrave 4. Nayor AR, Pollet JE: Eosinophilic colitis. Dis Colon Rectum EM, Germano D, Bravin J, McDonald T, Blair A, Adlard S, Jack- 1985; 615–618 son H: Randomized controlled trial of interventions designed to reduce the risk of progression to first-episode psychosis in a 5. Carpenter HA, Talley NJ: The importance of clinicopathological clinical sample with subthreshold symptoms. Arch Gen Psychi- correlation in the diagnosis of inflammatory conditions of the atry 2002; 59:921–928 colon: histologic patterns with clinical implications. Am J Gas- troenterol 2000; 95:878–896 2. McGlashan TH, Zipursky RB, Perkins DO, Addington J, Miller TJ, Woods SW, Hawkins KA, Hoffman R, Lindborg S, Tohen M, RAKESH KARMACHARYA, M.D., PH.D. Breier A: The PRIME North America randomized double-blind MARI MINO, M.D. clinical trial of olanzapine versus placebo in patients at risk of WILLIAM F. PIRL, M.D. being prodromally symptomatic for psychosis, I: study ratio- Boston, Mass. nale and design. Schizophr Res 2003; 61:7–18

JEAN ADDINGTON, PH.D. Clinical Trials During the Prodromal Stage DONALD ADDINGTON, M.D. of Schizophrenia Calgary, Alberta, Canada

TO THE EDITOR: With criteria that reflect an “ultra high risk” for developing a psychotic disorder, the risk of converting to psy- The Significance of Homocysteine Levels chosis is from 30% to 60% over 1 year (1, 2). To our knowledge, in Schizophrenia the first published trial on the subject (1) randomly assigned 59 ultra-high-risk subjects to 6 months of active treatment TO THE EDITOR: We noted with interest the findings of Donald (risperidone plus cognitive behavior therapy) or needs-based C. Goff, M.D., et al. (1) but question whether the comparison intervention. A second study, conducted by McGlashan and group they used was representative of homocysteine mea- colleagues from Yale and collaborating sites from North Caro- surement. Differences in age and homocysteine measure- lina, Toronto, and Calgary (2), used a randomized, double- ment and an absence of dietary history and genotyping for blind, placebo-controlled design with 60 prodromal subjects methylenetetrahydrofolate reductase, which are important

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1387 LETTERS TO THE EDITOR confounding variables, may have biased the homocysteine schizophrenia patients. This was not the intention of our analysis. study. Our hypothesis, that serum folate concentrations would The study group’s mean age was 20 years younger than the correlate inversely with the severity of negative symptoms, comparison group’s (the sixth Framingham Offspring cohort was based on the dual roles of glutamate carboxypeptidase II subgroup). Homocysteine levels rise progressively with age, as a modulator of brain N-methyl-D-aspartic acid activity and approximately doubling from childhood to old age. There is a as a facilitator of folate absorption in the intestines. This pri- bias toward higher comparison homocysteine levels as a re- mary hypothesis was supported by our results. Although a sult (2, 3). comparison of serum concentrations with a nonpsychiatric The accuracy of homocysteine differs among methods and study group was not an objective of our original study, we laboratories. This effect had not been controlled. The Framing- provided a comparison with the Framingham Offspring Study ham Offspring comparison group’s homocysteine level was cohort as a frame of reference for our findings. Dr. O’Donnell measured with high-performance liquid chromatography and Mr. Stephens are correct in stating that this comparison with fluorescent detection, whereas the study group used group was not matched for age (the mean age of our group fluorescence polarization immunoassay. The use of a local was 43 years versus 56 years for the Framingham group), nor comparison group with the same method and laboratory is were the same assay methods used. An ideal comparison recommended (3). group would be matched by age, smoking status, gender, ex- The Framingham Offspring Study shows that mandatory ercise level, and diet—a daunting task. We chose the Framing- fortification policy has reduced the prevalence of low folate ham Offspring Study sample because the subjects lived in the < status (i.e., 3.0 ng/ml) by more than 90% and the prevalence same geographical region, were sampled after folate supple- of mild fasting hyperhomocysteinemia (homocysteine con- mentation of grain products had been implemented, and > centrations 130 mmol/liter) by about 50% among its popula- were not taking vitamins with folate. The substantially lower tion-based cohort of middle-age to elderly U.S. citizens (4). folate concentrations both in our total sample and in schizo- We recommend that the assessment of homocysteine re- phrenia nonsmokers compared to this nonpsychiatric group quires the measurement of known confounding variables, support the hypothesis that low folate concentrations might especially in smaller group sizes, from folic-acid-fortified result from the low activity of glutamate carboxypeptidase II. regions. Contrary to larger previous studies with local com- As we emphasized in our discussion, there are several alterna- parison groups showing increased homocysteine levels in tive explanations for our findings. schizophrenia, these results should be treated with caution We did not design our study to test the hypothesis that ho- (5, 6). mocysteine concentrations are elevated in schizophrenia pa- References tients compared to a healthy population, and our results 1. Goff DC, Bottiglieri T, Arning E, Shih V, Freudenreich O, Evins should not be interpreted as such. We failed to find hypothe- AE, Henderson DC, Baer L, Coyle J: Folate, homocysteine, and sized clinical correlations with homocysteine serum concen- negative symptoms in schizophrenia. Am J Psychiatry 2004; trations except for a positive correlation with extrapyramidal 161:1705–1708 symptoms. Age did not predict homocysteine concentrations 2. Jacques PF, Selhub J, Bostom AG, Wilson PWF, Rosenberg IH: in our group, whereas gender was a very strong predictor that The effect of folic acid fortification on plasma folate and total should be controlled in future comparison studies. Homocys- homocysteine concentrations. N Engl J Med 1999; 340:1449– teine levels have varied widely among schizophrenia groups, 1454 having been found to be dramatically elevated primarily in 3. Refsum H, Smith AD, Ueland PM, Nexo E, Clarke R, McPartlin J, Johnston C, Engbaek F, Schneede J, McPartlin C, Scott JM: Facts young men in two groups from Israel (Applebaum et al., 2004, and recommendations about total homocysteine determina- and Levine et al., 2002) and not elevated in groups of schizo- tions: an expert opinion. Clin Chem 2004; 50:3–32 phrenia patients from the Netherlands (1) and Spain (2) and 4. Bostom AG, Selhub J, Jacques PF, Rosenberg IH: Power short- in female schizophrenia patients from Germany (3). We agree age: clinical trials testing the “homocysteine hypothesis” that careful examination of environmental and dietary factors against a background of folic acid-fortified cereal grain flour. will be important in understanding these findings. Clarifica- Ann Intern Med 2001; 135:133–137 tion of the possible neuropathological role of elevated ho- 5. Applebaum J, Shimon H, Sela BA, Belmaker RH, Levine J: Ho- mocysteine, as found by Levine and colleagues (2002), will be mocysteine levels in newly admitted schizophrenic patients. J of great interest to the field. Psychiatr Res 2004; 38:413–416 6. Levine J, Stahl Z, Ami B, Slava S, Ruderman GV, Belmaker RH: References Elevated homocysteine levels in young male patients with 1. Muntjewerff J-W, van der Put NM, Eskes T, Ellenbroek B, Stee- schizophrenia. Am J Psychiatry 2002; 159:1790–1792 gers E, Blom H, Zitman F: Homocysteine metabolism and B-vi- COLIN O’DONNELL, M.B.B.CH., M.R.C.PSYCH. tamins in schizophrenic patients: low plasma folate as a possi- TIM STEPHENS, B.SC. ble independent risk factor for schizophrenia. Psychiatry Res Melbourne, Victoria, Australia 2003; 121:1–9 2. Virgos C, Martorell L, Simo JM, Valero J, Figuera L, Joven J, La- Dr. Goff and Colleagues Reply bad A, Vilella E: Plasma homocysteine and the methylenetetrahydrofolate reductase C677T gene variant: lack of associa- TO THE EDITOR: We appreciate the letter from Dr. O’Donnell tion with schizophrenia. Neuroreport 1999; 10:2035–2038 and Mr. Stephens and agree that our study did not clarify the 3. Reif A, Schneider MF, Kamolz S, Pfuhlmann B: Homocysteine- question of whether homocysteine levels are elevated in mia in psychiatric disorders: association with dementia and

1388 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 LETTERS TO THE EDITOR

depression, but not schizophrenia in female patients. J Neural tion and function of dendritic spines. Proc Natl Acad Sci USA Transm 2003; 110:1401–1411 2000; 97:9287–9292 DONALD C. GOFF, M.D. 5. Fernandez A, Brautigan DL, Mumby M, Lamb NJ: Protein phos- Boston, Mass. phatase type-1, not type-2A, modulates actin microfilament integrity and myosin light chain phosphorylation in living non- muscle cells. J Cell Biol 1990; 111:103–112 Reduced Spinophilin in Schizophrenia 6. Grossman SD, Futter M, Snyder GL, Allen PB, Nairn AC, Green- TO THE EDITOR: Amanda J. Law, Ph.D., et al. (1) reported de- gard P, Hsieh-Wilson LC: Spinophilin is phosphorylated by creased mRNA for spinophilin in the hippocampal formation Ca2+/calmodulin-dependent protein kinase II resulting in reg- in patients with schizophrenia and mood disorders. The po- ulation of its binding to F-actin. J Neurochem 2004; 90:317– 324 tential implications of this abnormality are curious since spinophilin is involved in limiting or decreasing the length of 7. Rosoklija G, Toomayan G, Ellis SP, Keilp J, Mann JJ, Latov N, Hays AP, Dwork AJ: Structural abnormalities of subicular dendrites in dendritic spines. Spinophilin immunoreactivity is predomi- subjects with schizophrenia and mood disorders: preliminary nantly (although not exclusively) localized to dendritic spines findings. Arch Gen Psychiatry 2000; 57:349–356 (2), for which it is a meaningful marker, with hippocampal spine density and spinophilin immunoreactivity both in- ANDREW J. DWORK, M.D. creasing in response to estrogen (3). However, spinophilin is GORAZD ROSOKLIJA, M.D., PH.D. not necessary for the formation of spines, which are increased LIESL B. JONES, PH.D. in density in young spinophilin knockout mice or for their re- New York, N.Y. moval, since spine density returns to normal when such animals mature (4). Dr. Law and Colleagues Reply Spinophilin links protein phosphatase-1 to actin. Dephos- phorylation of actin by protein phosphatase-1 leads to its dis- TO THE EDITOR: Dr. Dwork and colleagues raise several inter- assembly (5), so the targeting of protein phosphatase-1 to ac- esting issues. We emphasized that our main conclusion was tin is likely to result in fewer or shorter dendritic spines. merely that the decrease in spinophilin mRNA provides a fur- Linking protein phosphatase-1 to actin also promotes de- ther indication that the site of molecular alterations in syn- phosphorylation of alpha-amino-3-hydroxy-5-methyl-4- apses in schizophrenia and mood disorders includes den- isoxazolepropionic acid receptors (4). The subsequent de- drites as well as presynaptic terminals. We also suggested, crease in the activity of alpha-amino-3-hydroxy-5-methyl-4- parsimoniously, that the reduction is probably related to the isoxazolepropionic acid receptors should decrease calcium decreases in spine density, which the correspondents and influx into the spine, reducing the phosphorylation of spino- others have clearly demonstrated. Given what is known of the philin by calcium-calmodulin kinase II and thereby enhanc- functions of spinophilin, as outlined in their letter, we agree ing its affinity for actin (6). that the reduction in its mRNA is more likely to be a conse- The reported subicular decrease in spinophilin mRNA in quence than a cause of the lower spine density (i.e., fewer schizophrenia is approximately 30% (1). We observed a de- spines require synthesis of less spinophilin), as we pointed crease of 75%–80% in the density of apical dendritic spines out in our article (p. 1862). The fact that the percentage decre- (plus a more modest decrease in apical dendritic arboriza- ment in spine density is much greater than the reduction of tion) (7), but the density of spines on basilar dendrites has not spinophilin mRNA can be explained, as Dr. Dwork and col- been measured. If the diminished pool of mRNA is support- leagues note, by the fact that changes in transcript abundance ing the synthesis of spinophilin for an even more diminished are often less than those of the encoded protein and that the set of dendritic spines, increased synthesis of spinophilin, on subcellular difference in the targeting of proteins may exist. a per-spine basis, might function to keep spine density re- Along similar lines, the lack of change in microtubule-associ- duced. Posttranscriptional mechanisms are also likely, as evi- ated protein 2 mRNA that we find may signify that any puta- denced by estrogen-induced increases in spine density and tive changes in microtubule-associated protein 2 are transla- spinophilin immunoreactivity, without altered levels of spino- tional or posttranslational in origin. philin mRNA (3). Dr. Dwork and colleagues proposed one cascade in which References spinophilin is involved and regulated. However, this important basic work was largely derived from model systems, and 1. Law AJ, Weickert CS, Hyde TM, Kleinman JE, Harrison PJ: Re- duced spinophilin but not microtubule-associated protein 2 the extent to which these molecular cascades are operating in expression in the hippocampal formation in schizophrenia the predicted manner in human hippocampal neurons is not and mood disorders: molecular evidence for a pathology of known. dendritic spines. Am J Psychiatry 2004; 161:1848–1855 In summary, we did not advocate that spinophilin is a mol- 2. Muly EC, Smith Y, Allen P, Greengard P: Subcellular distribution ecule with a specific or central etiological role in schizophre- of spinophilin immunolabeling in primate prefrontal cortex: nia or mood disorder, and we were unaware of the complexity localization to and within dendritic spines. J Comp Neurol of spinophilin’s roles in dendritic function. Rather, as the title 2004; 469:185–197 of the article stated, the aim was to complement and draw at- 3. Lee SJ, Romeo RD, Svenningsson P, Campomanes CR, Allen PB, Greengard P, McEwen BS: Estradiol affects spinophilin protein tention to the evidence that dendritic spines are part of the differently in gonadectomized males and females. Neuro- molecular and cellular neuropathology of schizophrenia and science 2004; 127:983–988 mood disorder—work at which Dr. Dwork and colleagues 4. Feng J, Yan Z, Ferreira A, Tomizawa K, Liauw JA, Zhuo M, Allen have been at the forefront. We are sure that they would agree PB, Ouimet CC, Greengard P: Spinophilin regulates the forma- that more studies are necessary to refine the evidence for

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1389 LETTERS TO THE EDITOR dendritic (spine) pathology and to begin to reveal the bio- and Proceedings Summary. Washington, DC, American Psychi- chemical pathways that underlie it. atric Association, 1982, number 95:275 3. Gold MS, Pomm R, Kennedy Y, Jacobs W, Frost-Pineda K: Five- AMANDA J. LAW, PH.D. year statewide study of physician addiction treatment out- CYNTHIA SHANNON WEICKERT, PH.D. comes confirmed by urine testing, in Proceedings of the 2001 THOMAS M. HYDE, M.D., PH.D. Meeting of the Society for Neuroscience. Washington, DC, Soci- JOEL E. KLEINMAN, M.D., PH.D. ety for Neuroscience, 2001, number 668.1 PAUL J. HARRISON, D.M., F.R.C.PSYCH. 4. Roy A: Suicide in doctors. Psychiatr Clin North Am 1985; 8:377– Oxford, U.K. 387 5. Simon W: Suicide among physicians: prevention and postven- Physician Suicide and Drug Abuse tion. Crisis 1986; 7:1–13 6. Gold MS, Byars JA, Frost-Pineda K: Occupational exposure and TO THE EDITOR: Physician suicide rates and suggestions for fu- addictions. Psychiatr Clin North Am 2004; 27:745–753 ture studies were nicely reviewed by Eva S. Schernhammer, 7. Gold MS, Dennis DM, Morey TE, Melker R: Exposure to narcotics M.D., Dr.P.H., and Graham A. Colditz, M.D., D.P.H. (1). How- in the operating room poses an occupational hazard for anes- ever, they might consider further evaluation of the “risk fac- thesiologists. Psychiatr Annals 2004; 34:794–797 tors relating to the working environment” (p. 2300). We have MARK S. GOLD, M.D. reported on outcomes of impaired physicians for nearly 25 KIMBERLY FROST-PINEDA, M.P.H. years and have followed all impaired Florida physicians since RICHARD J. MELKER, M.D., PH.D. 1995 (2, 3). Physician drug abuse has been linked to suicide (4, Gainesville, Fla. 5). We suggested that not all physician specialties are equally affected by drug abuse and dependence. Similarly, suicide may affect one medical specialty more than another. We have Dr. Schernhammer Replies suggested workplace evaluations, starting with a history of TO THE EDITOR: My associates and I appreciate the additional drug exposure in the operating and emergency rooms and in- background information conveyed by Dr. Gold and colleagues. tensive care units. Anesthesiologists are significantly overrep- Although presenting data on physician suicide rates by medi- resented among Florida physicians with substance use disor- cal specialty was beyond the scope of our analysis, we un- ders. They represent only 5.6% of the total licensed physicians doubtedly welcome their invitation for more research. but almost 25% of the physicians with substance use dis- Generally, men tend to successfully commit suicide more of- orders. Access to drugs of abuse has been the major theory ten than women, whereas suicide attempts among women are advanced to explain this. However, we have proposed that un- higher than those of men. It is possible, however, that women intended second-hand environmental exposure puts anes- physicians more successfully commit suicide than do women thesiologists at increased risk (6). We also recently demon- outside of the medical profession. Among U.S. medical stu- strated the presence of propofol and fentanyl in operating dents, the observed suicide rate of female students equaled room air after intravenous administration (7). Secondhand that of the male students (although they were still three to four exposure is a fact in some medical workplaces. It was not sur- times higher than those of their age mates), indicating a rela- prising that anesthesiologists and other physicians exposed tive scarcity of attempted suicides in that profession (1). to fentanyl in the workplace represented 90% of the fentanyl abusers in Florida. Studies in progress include sampling anes- That access to drugs can support higher suicide rates has long been shown: for example, in Australia, an increase in sui- thesiologists’ blood during work in cardiovascular surgery, an cides by women coincided with the implementation of a law environment where high doses of fentanyl are routinely used. that facilitated access to barbiturates (2). Therefore, it appears Environmental exposure may explain the high rates of addic- likely that higher suicide rates among physicians, who tend to tion among anesthesiologists and why recovery for anesthesi- prefer methods that are typical for their profession (3), may be ologists often necessitates giving up their work in operating coupled with both their easier access to drugs, as well as their rooms and even changing medical specialties. Prevention of better know-how concerning the successful use of such meth- physician opioid abuse and dependence appears to be linked ods. The interesting proposal by Dr. Gold et al. of unintended to identifying sources of secondhand exposure and prevent- secondhand environmental exposure to drugs as a risk factor ing exposure from occurring or by minimizing exposure, as for drug addiction and possibly suicide among anesthesiolo- was done with nitrous oxide. Environmental exposure may gists adds another layer of complexity and warrants further also prove to be an important factor in suicide attempts, re- investigation. lapses, and prevention. We would strongly suggest that new and important data from the analysis of Drs. Schernhammer References and Colditz be expanded to include medical subspecialty and 1. Pepitone-Arreola-Rockwell F, Rockwell D, Core N: Fifty-two secondhand exposure. medical student suicides. Am J Psychiatry 1981; 138:198–201 References 2. Oliver RG, Hetzel BS: Rise and fall of suicide rates in Australia: relation to sedative availability. Med J Aust 1972; 2:919–923 1. Schernhammer ES, Colditz GA: Suicide rates among physicians: 3. Bämayr A, Feuerlein W: Suizidhäufigkeit bei Ärzten und Zah- a quantitative and gender assessment (meta-analysis). Am J närzten in Oberbayern. Soc Psychiatry 1986; 21:39–48 Psychiatry 2004; 161:2295–2302 2. Gold MS, Extein I, Perzel J, Annitto WJ: Naltrexone in the treat- EVA SCHERNHAMMER, M.D., DR.P.H. ment of physician addicts, in 1982 Annual Meeting Syllabus Boston, Mass.

1390 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 LETTERS TO THE EDITOR

Methodological Concerns in a Trial 10 mg/day on days 3–7) and the maximum dose allowed (15 of Ziprasidone and Olanzapine mg/day). The authors’ argument that this dosing schedule is consistent with the package insert is weak. This dosing sched- TO THE EDITOR: In their randomized, double-blind trial compar- ule falls within the low range of the guidelines on the package ing ziprasidone and olanzapine for the treatment of acutely ill insert, which was based on studies performed about 10 years inpatients with schizophrenia or schizoaffective disorder, ago and submitted for approval to the Food and Drug Admin- George M. Simpson, M.D., et al. (1) provided important in- istration. Since that time, there have been many more studies formation showing that olanzapine-treated patients have a and a vast amount of clinical experience to suggest that a sig- greater risk of weight gain and lipid abnormalities than pa- nificantly faster upward titration rate and a higher maximum tients treated with ziprasidone. However, the dosing protocol dose often are needed to adequately treat acutely ill schizo- in this study raised a number of questions. First, there ap- phrenia patients (1). Therefore, the approach chosen by the peared to be a potential for unblinding. Each blister pack of authors was biased against finding efficacy in the olanzapine study medication was labeled “A,” “B,” or “C,” corresponding group. to a “low,” “medium,” or “high” dose of each drug. All ziprasi- Second, the vast majority of patients in both groups were done-treated patients were to receive the “high” dose at the treated with lorazepam in addition to the antipsychotic. Why end of 1 week, whereas the olanzapine-treated patients re- were such high rates of lorazepam allowed? Granted, acutely ceived the “medium” dose. During the trial, the treating clini- psychotic patients often benefit from a benzodiazepine. But cian would need to know the current dose classification each because a major goal of the study was to compare the efficacy week to decide whether it should or could be increased or de- of ziprasidone versus olanzapine, then the high rate of use of creased. A “medium” dose after the end of the first week would lorazepam limited the interpretation of the results. Benzo- clearly indicate olanzapine treatment, whereas a “high” dose diazepines help to enhance sleep and reduce agitation and would indicate ziprasidone treatment. It is possible that un- anxiety, thus interfering with the interpretations regarding ef- published procedures were used to prevent this potential ficacy of the antipsychotics. Furthermore, benzodiazepines problem. If so, knowledge of these procedures would be help- suppress antipsychotic-induced movement disorders such as ful in interpreting the results of the trial. akathisia, thus limiting the interpretation of the results re- A second concern with regard to the dosing protocol is one garding these adverse events. that is not uncommon in trials sponsored by pharmaceutical In summary, this was not simply a study of ziprasidone ver- companies, that of a suboptimal dose of a comparator drug. sus olanzapine; it was a study of ziprasidone plus lorazepam In this trial, the patients could receive a maximum olanzapine versus suboptimally dosed olanzapine plus lorazepam. dose of only 15 mg/day, although the product labeling recommended doses up to 20 mg/day (2). The patients received 10 Reference mg/day at the end of 1 week, and the mean dose of olanza- 1. Kinon BJ, Ahl JSV, Hill AL, Buckley PF: Dose response and atyp- pine throughout the trial was only 11.3 mg/day. In contrast, ical antipsychotics in schizophrenia. CNS Drugs 2004; 18:597– ziprasidone was titrated to the maximum dose recommended 616 by the product labeling (3), 160 mg/day, by the third day of the DAVID E. ROSS, M.D. trial. In order to reduce bias in studies comparing drugs of a Richmond, Va. sponsor and competitor, available doses should include the entire range recommended by the product labeling. Dr. Simpson Replies

References TO THE EDITOR: I thank Drs. Carnahan, Perry, and Ross for their 1. Simpson GM, Glick ID, Weiden PJ, Romano SJ, Siu CO: Random- comments on our study and welcome the opportunity to dis- ized, controlled, double-blind multicenter comparison of the cuss the issues they raise regarding drug administration, con- efficacy and tolerability of ziprasidone and olanzapine in comitant medication, and dosing. acutely ill inpatients with schizophrenia or schizoaffective dis- Drs. Carnahan and Perry comment that the medication la- order. Am J Psychiatry 2004; 161:1837–1847 beling—“A,” “B,” and “C,” denoting “low,” “medium,” and 2. Package insert for Zyprexa. Indianapolis, Eli Lilly and Company, “high” doses—would alert investigators at day 7 to treatment 2001 3. Package insert for Geodon. New York, Pfizer Inc, 2004 assignment because all patients randomly assigned to ziprasidone received 80 mg b.i.d. (the “high” dose) on days 3 to 7 and RYAN M. CARNAHAN, PHARM.D., M.S. all patients randomly assigned to olanzapine received 10 mg/ Tulsa, Okla. PAUL J. PERRY, PH.D. day (the “medium” dose). The “A,” “B,” and “C” labeling was, Iowa City, Iowa in fact, employed only during the flexible-dose weeks of the study (weeks 2 to 6). During days 1 to 2 and days 3 to 7, when fixed doses were administered, the medication cards did not contain this labeling. During both the fixed titration and flex- TO THE EDITOR: I read with interest the study by Dr. Simpson ible-dose phases of the study, the patients in the two treat- and colleagues comparing ziprasidone to olanzapine for the ment arms received identical quantities of medication of treatment of schizophrenia. I commend the authors for com- identical appearance; in the olanzapine group, placebo cap- pleting a head-to-head study that provides some important sules were employed to simulate twice-a-day dosing. For ex- safety data. However, the study has two important limitations ample, a subject who was assigned to olanzapine at 10 mg/ that limit its interpretation. day would have received the same number of identical-ap- First, the olanzapine dosing was excessively restricted with pearing capsules twice a day as a subject who was assigned to respect to the upward titration rate (5 mg/day on days 1 and 2; ziprasidone at 80 mg b.i.d. with this “double-dummy” design.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1391 LETTERS TO THE EDITOR

Thus, there was no potential for unblinding in the drug-ad- results of the North American double-blind olanzapine trial. ministration protocol. Neuropsychopharmacology 1996; 14:111–123 Dr. Ross questions the use of lorazepam in our trial to treat 3. Beasley CM Jr, Hamilton SH, Crawford AM, Dellva MA, Tollefson GD, Tran PV, Blin O, Beuzen JN: Olanzapine versus haloperidol: agitation or insomnia. Clinical trials of antipsychotics in acute phase results of the international double-blind olanza- acute schizophrenia have commonly permitted concomitant pine trial. Eur Neuropsychopharmacol 1997; 7:125–137 use of benzodiazepines, although data on their use have not 4. Tollefson GD, Beasley CM Jr, Tran PV, Street JS, Krueger JA, always been reported. The percentages we reported in our ar- Tamura RN, Graffeo KA, Thieme ME: Olanzapine versus halo- ticle for any use of lorazepam during the study—83.1% in the peridol in the treatment of schizophrenia and schizoaffective ziprasidone group and 75.2% in the olanzapine group—re- and schizophreniform disorders: results of an international flected lorazepam use during days 1 to 7. During days 8 to 14, collaborative trial. Am J Psychiatry 1997; 154:457–465 the percentages of subjects still taking lorazepam decreased 5. Conley RR, Mahmoud R: A randomized double-blind study of to 51.5% in the ziprasidone group and 51.9% in the olanza- risperidone and olanzapine in the treatment of schizophrenia pine group. Lorazepam use continued to decrease over subse- or schizoaffective disorder. Am J Psychiatry 2001; 158:756– quent study weeks and remained comparable for the two 774; correction, 158:1759 treatment groups throughout the study. GEORGE M. SIMPSON, M.D. With regard to the comments on dosing, the olanzapine Los Angeles, Calif. dosing regimen employed in the trial was consistent with dosing recommendations from the olanzapine prescribing infor- Olanzapine and Haloperidol mation, as well as with published clinical trial data available for Residual Symptoms at the time the study was designed and conducted (1–4). It should be noted that only olanzapine-naive patients were in- TO THE EDITOR: Robert W. Buchanan, M.D., and colleagues (1) cluded in our study. Patients who had more than 14 days of to- conducted an important study on the comparative effect of tal lifetime exposure to olanzapine or who had received an olanzapine and haloperidol on residual positive and negative olanzapine dose above 10 mg/day were excluded. This crite- symptoms in 63 outpatients with treatment-resistant schizo- rion for entry, together with the fact that patients who are re- phrenia. They concluded that olanzapine has limited differ- quired to provide informed consent in such clinical trials are ential benefit for positive and negative symptoms in these pa- generally not of the highest level of illness severity, may have tients. However, we think that the correct conclusion must be moderated the average dosing requirements for subjects in that olanzapine has no benefit over haloperidol. Namely, the the study. The mean olanzapine dose in our study reached 13 authors could not find superior efficacy for olanzapine for mg/day by days 15 to 21 and remained at that level for the re- positive and negative symptoms, and the main side effects maining study weeks (the mean dose for the entire length of (extrapyramidal symptoms for haloperidol and weight gain the study was 11.3 mg/day), indicating that this medication for olanzapine) seem to balance each other out. The authors was adequately dosed. As noted in our article, the mean olan- stated that “the magnitude of weight gain…may potentially zapine dose after week 5 in our study—13.1 mg/day—is virtu- offset the more benign extrapyramidal symptom profile of ally identical to the endpoint mean dose in the comparative olanzapine” (p. 128). Furthermore, we would like to point out trial of olanzapine and risperidone reported by Conley and that the mean doses of olanzapine (20.3 mg/day) and halo- Mahmoud (5). peridol (18.3 mg/day) resulted in noncomparable dopamine D receptor occupancy. For almost all patients receiving a Although ziprasidone could be administered at up to 160 2 dose of about 18 mg/day of haloperidol, the balance between mg/day, the maximum dose currently recommended in the efficacy and extrapyramidal side effects is not optimal (2, 3). product labeling (the package insert for Geodon), the mean Haloperidol at lower doses is thought to induce fewer ex- doses were approximately 140 mg/day at days 15 to 21 and trapyramidal side effects and fewer neuroleptic-induced neg- the following weeks, with a mean dose of 130 mg/day for the ative symptoms and dysphoria (4) without a change in effi- entire length of the study, indicating that many subjects were cacy on positive symptoms (2). Lower haloperidol doses judged not to require the maximum recommended dose. The might even further diminish the benefit of olanzapine for ex- percentages of subjects receiving the maximum dose at end- trapyramidal symptoms. point were similar for the olanzapine (59.4%) and ziprasidone Therefore, the sobering conclusion of this study seems to (61.8%) groups. In addition to observing comparable efficacy be that neither olanzapine nor haloperidol shows benefits for between ziprasidone and olanzapine in the treatment of hos- outpatients with schizophrenia who met criteria for either re- pitalized patients with acute schizophrenia, we found differ- sidual positive or residual negative symptoms. ences between treatment groups in body weight, lipid profile, and insulin resistance. Presumably, higher doses would not References have ameliorated—and may have exacerbated—these ad- 1. Buchanan RW, Ball MP, Weiner E, Kirkpatrick B, Gold JM, McMa- verse events observed in the olanzapine-treated subjects. hon RP, Carpenter WT Jr: Olanzapine treatment of residual positive and negative symptoms. Am J Psychiatry 2005; 162:124– References 129 1. Beasley CM Jr, Sanger T, Satterlee W, Tollefson G, Tran P, Hamil- 2. Bollini P, Pampallona S, Orza MJ, Adams ME, Chalmers TC: An- ton S: Olanzapine versus placebo: results of a double-blind, tipsychotic drugs: is more worse? a meta-analysis of the pub- fixed-dose olanzapine trial. Psychopharmacology (Berl) 1996; lished randomised control trials. Psychol Med 1994, 24:307– 124:159–167 316 2. Beasley CM Jr, Tollefson G, Tran P, Satterlee W, Sanger T, Hamil- 3. Kapur SJ, Zipursky R, Jones C, Remington G, Houle S: Relation- ton S: Olanzapine versus placebo and haloperidol: acute phase ship between dopamine D2 occupancy, clinical response, and

1392 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 LETTERS TO THE EDITOR

side effects: a double-blind PET study of first-episode schizo- 2. Tollefson GD, Birkett MA, Kiesler GM, Wood AJ (Lilly Resistant phrenia. Am J Psychiatry 2000; 157:514–520 Schizophrenia Study Group): Double-blind comparison of olan- 4. de Haan L, van Bruggen M, Lavalaye J, Booij J, Dingemans zapine versus clozapine in schizophrenic patients clinically eli- PMAJ, Linszen D: Subjective experience and D2 receptor occu- gible for treatment with clozapine. Biol Psychiatry 2001; 49: pancy in patients with recent-onset schizophrenia treated with 52–63 low-dose olanzapine or haloperidol: a randomized, double- ROBERT W. BUCHANAN, M.D. blind study. Am J Psychiatry 2003; 160:303–309 M. PATRICIA BALL, R.N. LIEUWE DE HAAN, M.D., PH.D. ELAINE WEINER, M.D. Amsterdam, the Netherlands Baltimore, Md. NICO VAN BEVEREN, M.D. Rotterdam, the Netherlands Origin of the Term “Schizophrenia” Dr. Buchanan and Colleagues Reply TO THE EDITOR: The portrayal by Ernest L. Abel, Ph.D., of the 1857 “theory of degeneration” of Benedict-Augustin Morel (1) TO THE EDITOR: Drs. de Haan and van Beveren raise two points as a “parsimonious explanation for the etiology of insanity” concerning our recent article on the comparative efficacy of dominating French psychiatry for “almost a century” overes- olanzapine and haloperidol for residual positive or negative timated its influence by almost half a century. symptoms. First, they note that in the abstract of our article By 1911, Swiss psychiatrist Paul Eugen Bleuler had renamed we stated that “Olanzapine has limited differential benefit for Kraepelin’s 1899 Latin form of Morel’s earlier term demence either positive or negative symptoms in patients with treat- precoce, “schizophrenia,” emphasizing that the illness known ment-resistant schizophrenia” (p. 124). They suggest that the as “dementia praecox” was not an actual dementia and did proper conclusion should be that “olanzapine has no benefit not always begin at an early age. Although the growing influ- over haloperidol.” The qualification in our statement reflects ence of German psychiatry in France induced negative reac- the fact that our study is not the only one to have addressed tions, whose main target was the work of Kraepelin, the na- this issue. In our Discussion section, we reviewed two studies tionalistic tone reached its apex on the French side during that asserted a benefit for olanzapine in this population (1, 2). World War I, and only some faint traces remained for several Although we believe that these studies have serious method- years after 1920 (2). ological flaws, they remain in contradistinction to our results. Moreover, the 1950 discovery of chlorpromazine by physi- Second, Drs. de Haan and van Beveren suggest that the mean cian Henri Laborit and the seminal work in 1952 by French doses of haloperidol and olanzapine achieved in our study psychiatrists Jean Delay and Pierre Deniker, which intro- result in noncomparable D2 receptor occupancy, with the duced the chemical as a treatment for schizophrenia (3), haloperidol dose associated with increased D2 receptor oc- demonstrated French psychiatrists’ much earlier acceptance cupancy, which could potentially lead to increased extra- of the Bleuler model. pyramidal symptoms, dysphoria, and secondary negative Although today we can look to Morel’s work as a progenitor symptoms. Although we agree with the theoretical concern, of the current biological approach to psychiatric illnesses, we note that the haloperidol-treated patients did not exhibit a Bleuler’s characterization—“the group of schizophrenias”— mean worsening of either extrapyramidal symptoms or de- remains the more parsimonious approach. pressive symptoms (as measured by the Hamilton Depression Scale) nor a worsening of negative symptoms (Tables 2 and 3). References The alternative concern, which we raised in our Discussion 1. Abel EL: Benedict-Augustin Morel (1809–1873) (image, psych). section, is that the olanzapine dose was relatively low for this Am J Psychiatry 2004; 161:2185 population. A higher olanzapine dose may have led to in- 2. Pichot P: [The history of German psychiatry from the viewpoint of French psychiatrists]. Fortschr Neurol Psychiatr 1992; 60: creased symptomatic improvement. 317–328 (German) References 3. Etain B, Roubaud L: Jean Delay, MD, 1907–1987 (image, psych). 1. Breier A, Hamilton SH: Comparative efficacy of olanzapine and Am J Psychiatry 2002; 159:1489 haloperidol for patients with treatment-resistant schizophre- JACK C. SCHOENHOLTZ, M.D. nia. Biol Psychiatry 1999; 45:403–411 Mamaroneck, N.Y.

Reprints are not available; however, Letters to the Editor can be downloaded at http://ajp.psychiatryonline.org.

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1393 Book Forum

NEUROLOGY AND NEUROPSYCHIATRY The Neurological Basis of Pain provides the latest thinking and practice in the field and helps the reader grasp much of the vast biology of neuroplasticity, the basis of neuropathic pain. Focusing on symptoms and etiology has not provided a The Neurological Basis of Pain, edited by Marco Pappa- method of intervention, however; similar symptoms may arise gallo, M.D. New York, McGraw-Hill, 2005, 840 pp., $159.00. from different etiologies, and the same etiology may be associated with a range of symptoms. Thus, research is progressing McGraw-Hill invited Dr. Pappagallo to edit this book as a to describe the underlying mechanisms and the means by companion to Adams & Victor’s Principles of Neurology (1). which they can be clinically identified and managed. Dr. Por- Dr. Pappagallo set out “to create a text that would command tenoy, in the foreword, states that the development of new the respect of neurologists.” He also hoped “a variety” of other drugs and other treatments “holds the promise of true mecha- “professionals” would “benefit from it.” He has achieved his nism-based therapy.” Thus, the wheel may turn again and the goals and deserves congratulations. biopsychosocial approach may become redundant. The book is arranged in four parts dealing, respectively, There is a dearth of psychiatrists among the authors, but with the scientific basis of pain, physician-patient issues, the psychologists are well represented. This reflects real-world syndromes and disorders, and, finally, the therapies. Natu- pain management, although psychiatrists have much to offer rally, when multiple authors deal with overlapping topics, this field, and the work is rewarding. In a chapter on spinal there is some repetition, but it is only of important points and pain, Wheeler and Murrey state that training should be made does not distract from this monumental work. available to neurologists and psychiatrists so they can estab- There are 46 chapters by 76 authors. Fifty-four are medi- lish “competence in using electrodiagnostic skills for diagno- cally trained, and 20 of these appear to be attached to neurol- sis and performing spinal interventional treatments with spi- ogy and 13 to anesthesiology departments. About nine appear nal injections.” I hope this comes to pass. to be surgeons, and two seem to be psychiatrists. The Neurological Basis of Pain is a magnificent contribu- The culture, practice, and structure of medicine is rapidly tion to pain management. It has the combination of readabil- changing. Pain medicine is a prime example. As Dr. Portenoy ity and scholarship that will make it a classic. It should be in points out in the foreword, pain was once considered as a every medical library and on the shelf of everyone with a seri- symptom of disease but has recently been conceptualized as ous interest in pain medicine. a disease itself. Thus, there are many applicants for this es- References sentially vacant property. The Neurological Basis of Pain 1. Victor M, Ropper AH: Adams & Victor’s Principles of Neurology, stakes a claim for the neurologists. In much of the United 7th ed. New York, McGraw-Hill, 2000 Kingdom, Europe, and Australia, the anesthetists are the main 2. Engel GL: The need for a new medical model: a challenge for contenders, but pain management is achieving adulthood biomedicine. Science 1977; 196:129–136 and is in the process separating. About seven authors of this SAXBY PRIDMORE, M.D. book list their attachment to a pain management unit, with Hobart, Tasmania, Australia no hint of subservience to neurology, anesthesiology, or other parental discipline. Neurology for the Non-Neurologist, 5th ed., by William During the maturation period of the last 50 years, pain Weiner and Christopher Goetz. Philadelphia, Lippincott Wil- medicine has been giving two messages: 1) pain needs spe- liams & Wilkins, 2004, 528 pp., $59.95 (paper). cial, separate consideration, and 2) pain is complex and requires contributions from diverse fields, not only neurology, This multiauthored book provides a practical overview of anesthesiology, and surgery but also pharmacology, psychol- clinical neurology for physicians who work in other areas of ogy, and physical therapy, among others. medicine. It begins with a concise and well-focused discus- The multidisciplinary approach of the book gives a voice to sion of the basic neurological examination, beginning with all. The only irritating chapter (which deals with physical the cranial nerves and peripheral nervous system and moving therapy) rails against “traditional-thinking clinicians en- systematically through basic diagnostic tests and the evalua- trenched in a narrow, structurally focused biomedical (im- tion of coma. Each chapter is individually authored and cov- pairment) model” and recommends a “patient-centered bio- ers a broad variety of neurological problems, including cere- psychosocial model.” This is a bit hard to take because it is brovascular disease, headache, epilepsy, multiple sclerosis, many decades since Dr. George Engel (2) provided the con- Parkinson’s disease, alcoholism, peripheral neuropathy, the cept and label of the biopsychosocial approach, in which dementias, traumatic brain injury, neuromuscular disorders, medicine is now well marinated. and neurological manifestations of cancer. There is also an In chapter 15, Hord and colleagues state, “Chronic pain is entire chapter devoted to neurological complications of preg- the result of plastic changes in the peripheral and/or central nancy. These topics are all dealt with in a straightforward way nervous systems induced by injury or dysfunction in the ner- that is suitable for the average psychiatrist. NeuroAIDS and vous system.” Although this is a generalization and there is lit- other conditions with relevance to psychiatry, such as low tle doubt that inflammation can result in chronic pain rela- back pain and insomnia, are also dealt with in a clear and tively independent of change within the nervous system, it is practical way. probably correct that most of the patients treated in chronic Most chapters provide a description of the correct ways to pain units have neuropathic pain. perform a neurological examination and enumerate the

1394 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BOOK FORUM forms of treatment that are available for different neurologi- any good core textbook would already cover these areas and cal disorders. In the chapter on “eye signs,” detailed diagrams more besides. More senior colleagues will not find the defini- showing visual pathways are provided and visual field defects tive, up-to-date reviews that they would otherwise look for in are depicted in simple diagrams. The final chapter includes a bigger textbooks or journals. I could not think of a particular discussion of medical-legal issues involved in the care of pa- group of psychiatrists, psychologists, or neurologists to whom tients with neurological disorders as well as a discussion of I would recommend this book. There will be future titles in the problems associated with obtaining informed consent this series, and my old teacher would perhaps have written, from patients who are not legally competent. This chapter “We are hoping to see some of the promised talent in the next also contains a fairly extensive discussion of brain death, in- term.” cluding the use of electroencephalography and cerebral ROBERT HOWARD, M.D., M.R.C.PSYCH. blood flow as part of the evaluation. For the psychiatrist who London, U.K. is seeking to learn more about behavioral neurology, there is a description of the Papez circuit and a fairly comprehensive discussion of temporal lobe epilepsy. Beyond this, however, there is very little information regarding psychiatric manifes- SLEEP DISORDERS tations of neurological disorders. For those who treat the geri- atric population, where behavioral manifestations of delirium and dementia are common, this book is probably not an Sleep Medicine in Clinical Practice, by Michael H. Silber, appropriate resource. Overall, however, Neurology for the M.B.Ch.B., F.C.P.(SA), Lois E. Krahn, M.D., and Timothy I. Mor- Non-Neurologist provides a well-illustrated and clearly writ- genthaler, M.D. New York, Taylor & Francis (Mayo Foundation ten overview of basic clinical neurology. for Medical Education and Research), 2004, 392 pp., $99.95. FRANCINE M. BENES, M.D., PH.D. Belmont, Mass. The study and treatment of sleep disorders is a young, rapidly expanding specialty, with board certification, accredited treatment centers, and research journals. Combining aspects Neuropsychiatric Assessment, edited by Stuart C. Yudof- of psychiatry, neurology, and pulmonary medicine, it deals sky, M.D., and H. Florence Kim, M.D. Washington, D.C., Amer- with an important human function largely ignored until the ican Psychiatric Publishing, 2004, 208 pp., $34.95 (paper). discovery of rapid eye movement (REM) sleep (1). Half a cen- I was hopelessly bad at mathematics at school and was tury later, hundreds of sleep laboratories offer diagnostic stud- happy to leave a world of half-understood quadratic equa- ies, and scores of accredited sleep disorder centers provide tions and integral calculus well and truly behind me in my both diagnosis and ongoing treatment for insomnia, obstruc- teens. Yet, a teacher’s end-of-term report on my progress in tive sleep apnea, narcolepsy, and other sleep-related illnesses. the subject has stuck with me as a cutting, if accurate, de- The authors’ stated audience for this compact volume in- scription of more generalized ability: “Tries hard—but then cludes trainees, practicing sleep physicians, and academic he has to.” This reassured my parents that I was at least doing training programs, and they hope to spur interest in those my best and gave me a useful insight into my personal limita- outside the field. They reflect their own practice at the Mayo tions. Many years older and not much wiser, I appreciate how Clinic; for example, they describe a night in their sleep labora- much harder it is to write an honest—but essentially critical— tory that differs in some interesting details from the proce- report than one in which positive gloss, thickly applied, ob- dures of other accredited centers. Their overview is organized scures the message. into four sections: Basics of Sleep Medicine, The Sleepy Pa- In their introduction, the editors tell us that this volume in tient, The Patient Who Cannot Sleep, and The Patient With the Review of Psychiatry series provides an overview of the Excessive Movement in Sleep. In this general approach lie neuropsychiatric approach to assessment and presents dis- both the strengths of this book and its weakness. cussions of techniques and testing methods that may be more As a short textbook, Sleep Medicine in Clinical Practice is a familiar to neurologists than to psychiatrists. The book cer- useful addition to the available general literature, suitable both tainly does what it claims to on the cover: five chapters review as an overall introduction and as a quick reference for busy the neuropsychiatric and neuropsychological examinations, practitioners. It covers the diagnostic categories within the cur- electrophysiological and pathological laboratory testing, and rent (soon to be revised) International Classification of Sleep neuroimaging. Disorders, some very briefly. Clinical norms, often adjusted by The book is attractively produced, and when I had finished age group, will help the general physician interpret sleep study reading the introduction to the Review of Psychiatry series reports. Illustrative clinical vignettes are well placed and help- that occupied the first few pages, I felt excited by the prospect ful, fleshing out the diagnostic entities. Charts, tables, and line of what lay ahead. I am very sorry to say that I was disap- drawings illustrate key topics and provide examples of poly- pointed by all five of the book’s main chapters. None of them somnographic studies. Useful clinical algorithms show how to really offers much more than a beginner’s guide to the sub- reach clinical diagnoses and treatment decisions. ject, and some are woefully patchy in their coverage of chosen The authors’ effort to provide a wide overview of sleep med- areas. The more I read, the more puzzled I became as to icine, however, ultimately limits the book’s usefulness. Rare exactly who the target readership for this book might be. and unusual diagnoses are covered, but the same succinct ap- Medical students or perhaps the most junior of psychiatric or proach is applied to common disorders. For example, the au- neurological trainees will find competent accounts of neu- thors describe recurrent (periodic) hypersomnia and catathre- ropsychological and neuropsychiatric assessments here, but nia (nocturnal expiratory groaning), illnesses the average

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1395 BOOK FORUM clinician will probably never encounter. Sleep apnea, both ob- this organization. For example, Peled sees dysthymia as “recur- structive and central, makes up the bulk of clinical work, com- rent deoptimization shifts of the transmodal levels accompa- prising perhaps four of every five referrals to a sleep center, but nied by constraint frustration” (p. 100), bipolar mood swings as receives only 50 pages of compact description, roughly one- an “oscillatory dynamic of optimizations and deoptimizations” eighth of the book. Restless legs syndrome and periodic limb (p. 72), psychosis as “connectivity breakdown of the dynamic movement disorder, two other common entities, receive even core” (p. 72), anxious loss of control as “destabilization of the less space. Insomnia is better covered, with separate chapters higher level transmodal brain systems relevant to conscious devoted to causes, clinical approach, and management, but it awareness” (p. 73), and transference as “activation of the at- too receives less attention than its high incidence merits: 17 tractor systems which represent the person from the past” (p. pages. The coverage of delayed sleep phase and other disor- 75). He does not address character types, which I believe can be ders of circadian rhythm is also thin. modeled plausibly by tweaking neural network elements. For a reference source, this book’s index is only adequate. Peled provides illustrative cases for his system of psychiatric For example, although tobacco use and its effect on sleep are brain profiling in which both “external and internal perturba- mentioned in several chapters, the index has no listing for to- tors affect the system development and organization” (p. 93). bacco, smoking, or nicotine, and nicotine use is not included For external perturbators he favors the Holmes-Rahe Social in the discussion of substance abuse and sleep. Cited refer- Adjustment Scale (p. 94), which ranges from minor violations ences are listed at the end of chapters, but for those seeking of the law and Christmas to divorce and death of spouse. Inter- additional sources a bibliography would be useful, and this nal perturbators include metabolic, medication, and intracra- book has none. nial pathological effects. Peled says his Psychiatric Brain Pro- In short, Sleep Medicine in Clinical Practice is a useful intro- file is less stigmatizing and categorical and has more “degrees duction to an exciting medical field, but readers looking for of freedom” than DSM and yet is more constrained by neuro- greater depth of clinical information will need to go elsewhere. science than psychoanalytic conceptualizations that “have unlimited degrees of freedom allowing for all concepts to de- Reference scribe all occurrences and thus are operationally meaning- 1. Aserinsky E, Kleitman N: Regularly occurring periods of eye less” (p. 95). Many of Peled’s sources as well as his subtitle, motility, and concomitant phenomena, during sleep. Science however, are conceptually derived from psychoanalysis. 1953; 118:273–274 The book concludes with ideas about future directions for RICHARD B. MAKOVER, M.D. psychiatry. Testable hypotheses must move from a linear two- New Haven, Conn. factor model of cause and effect to a three-level model of lower-level multiple biological factors, an intermediate-level system organization, and higher-level system functions and MODELS OF THE MIND emergent functions. Means to detect perturbations of brain organization must be developed. Synaptogenic control should include neurogenesis. Direct pacemaker control could include transcranial magnetic stimulation coupled with EEG Brain Dynamics and Mental Disorders: Project for a as well as imaging and deep brain stimulation. Experience Scientific Psychiatry, by Avi Peled, M.D. Tel Aviv, Israel, Yoz- control should include virtual reality technology, which Peled mot Heiliger Publishers, 2004, 121 pp., $24.00 (paper). believes has potential to correct specific brain cognitive defi- ciencies, even delusions (for example, by showing patients Toward the end of the 1980s neural network theories of with delusions of persecution by the FBI a warm and caring connectionism became sufficiently current that we began to feel that, yes, for the first time we could see how the brain FBI headquarters). might self-organize and generate mental experience. This If the brain were a corporate office, our present state of brief book reviews the recent progress in this newly dominant functional imaging would put us in the position of the super- way of thinking about the logical-computational basis of intendent in the basement who can monitor departments’ brain function and proposes a framework for conceptualizing use of electrical power and tell who is burning the midnight psychopathology. Readers who find the author’s clear exposi- oil. Peled seems to be proposing that psychiatrists assume the tion a bit spare might wish to consult some of his most heavily role of a corporate information technology manager who relied upon sources, such as Hebb, Hopfield, Rumelhart and knows the information storage and transmission require- McClelland, Hoffman, Goldman-Rakic, Tononi, Edelman, ments of each department as well as the volume and destina- and Mesulam, especially if such terms as “Hebbian plasticity,” tions of its e-mail and who addresses bugs, overloads, “Hopfield nets,” “parallel distributed processing,” “pathologi- crashes, and viruses that arise in the system. In a continuation cal foci,” “reverberating network feedback,” “reentry,” “neural of the metaphor, psychoanalysis would be a little like entering Darwinism,” and “heteromodality” are not words found lying the play sphere of the office party to observe the employees’ about their households. interactions, or perhaps taking the chief executive officer’s The subtitle echoes Freud’s “Project for a Scientific Psychol- secretary to lunch to hear gossip about office politics. We still ogy” (1) and attempts to continue his abandoned, premature do not know how to read the e-mail and must infer how the effort to explain neuronal roots of mind. Mental functions can corporation does its business, decides its priorities and strat- now be viewed as emergent properties of complex brain orga- egies, innovates, integrates its employees’ expertise, and sets nization, and mental disorders can be seen as perturbations of departmental budgets.

1396 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BOOK FORUM

Reference The Philosophy of Psychiatry: A Companion, edited by 1. Freud S: Project for a scientific psychology (1950 [1887–1902]), Jennifer Radden. New York, Oxford University Press, 2004, 447 in Complete Psychological Works, standard ed, vol 1. London, pp., $74.00. Hogarth Press, 1966, pp 295–397 The 30 chapters of this book are written by 25 philosophers DAVID V. FORREST, M.D. and a handful of clinicians. Most are written by a single au- New York, N.Y. thor, addressing the many philosophical aspects of psychiatric practice and targeting issues such as what it means to The Mind: Its Nature and Origin, by Christiaan D. van make a diagnosis, when is one not responsible for a criminal der Velde, M.D. Amherst, N.Y., Prometheus Books, 2004, 242 act, what does competence imply, and when is the self contin- pp., $28.00. uous or discontinuous. There are philosophical discussions of such varied topics as brain pain, desire, memory, values, evo- Given the expansive title of this book, I expected to find a lution, research ethics, religion, race, and gender. cryptic and complex tome of many hundreds of pages filled I wish I could say that the promise inherent in the nature of with diagrams, positron emission tomography scans, and the the topics translates into great reading, but it doesn’t. The most modern three-dimensional images. As it turned out, I idea for the book is excellent, and this may prove a very suc- was partially correct. The Mind does seem cryptic and com- cessful text if the intended readers are philosophy students. plex to me, but it runs under 250 smallish pages and fulfills Clinicians, however, will find most of the chapters hard to the promise implied about the nature or origin of one’s mind. read. The language of academic philosophy is not the lan- I am willing to accept some responsibility for not fully guage of bedside psychiatry. With some exceptions, the sin- grasping the merits of the book. Perhaps I am too narrowly fo- gle-authored chapters by philosophers are so dense, so laden cused in biologically oriented psychiatric medicine to appre- with jargon, and so embedded in a philosophical context in- ciate this effort. However, some fault must lie with the author scrutable to the ordinary reader that their message is lost. and editors for producing a book whose book jacket sum- A notable exception is the chapter by Jennifer Church on mary, a reflection of the contents, is so convoluted that it left the social construction of madness. This is a wonderful explo- me and a few colleagues puzzled. ration of the pros and cons of viewing illness as socially con- The quest, we are told, is to demonstrate how cerebral ac- structed versus accepting it as biologically predetermined. tivities become mental events. On this journey of 21 chapters This chapter is insightful, thought-provoking, and written in organized within five parts we revisit Freud, Piaget, and many plain English. Another very readable, balanced, and useful other revered psychologists and behavioral scientists; en- chapter written by a philosopher, Daniel Robinson, is on the grams; and Gestalt. We are treated to the history of dialectical concept of dangerousness. On the whole, the chapters that concepts, beginning with Zeno of Elea in 464 BCE. We are of- work best are those written by a philosophy/psychiatry team. fered dialectical interconnectedness and dialectical triads. Sadly, there aren’t many of these. Michael Schwartz and Os- We are awash with matrices and fusions as well as complex borne Wiggins contribute a very good chapter dispelling the and simple mnemonics. Part 3, Structures of the Mind, has a myth that clinical drug trials and neuroscience constitute the sole scientific methodologies of psychiatry. They talk about chapter titled “Ego, Superego, Id,” and the only other chapter understanding and interpretation, the methodology of study- is titled “The Neurophysiology of Dreaming.” ing psychopathology and psychotherapy. This is important The section on Brain, Mind and Body includes six chapters because it broadens the focus of what psychiatric practice en- totaling 28 pages. The chapter on the “Mind-Body Problem” is compasses and shields it from allegations of reductionism, less than four pages long and has three references, from 1950, charges to which biomedicine is vulnerable. Most of the co- 1980, and 1985 (by the author), and includes a figure (number authored chapters are worth reading—a comment on the fact 10) of the ubiquitous smiley face and sad face. that when two disciplines work together to produce a piece of The section on Psychological and Clinical Implications writing, they abandon the jargon of their respective fields and covers personality formation, psychopathology, and psycho- write in a style that others can understand. therapy in a brief 40 pages. Although it might fortify the previ- Because the idea for a book of this kind is so good, I would ous chapters, as a clinician I found that it did not offer much encourage the editor to try again. My prescription for excel- to strengthen my practice or pearls to share with residents. lence would be to restrict the number of topics and to select Pointing out what I see as flaws in this book does not neces- authors carefully. I would insist that philosophers and psychi- sarily mean I found no merit. I look at The Mind as primarily a atrists collaborate on each chapter. I would not allow bland relatively brief philosophical, psychological, and scientific ex- reviews of a topic area and would insist that each chapter de- ploration of concepts. For those wanting to ponder “the fend a point of view. I would ensure that all chapters be read mind,” memory, and cognition relative to a historical frame- by all authors and that wrestling with each other’s arguments work it may well be a very rewarding experience. For those be part of the task of writing. I would not permit the use of vo- looking for a more useful clinical or teaching tool for psychi- cabulary that the general educated public does not under- atric medicine I do not think this would be a first-line choice. stand. I look forward excitedly to such a book. MARK H. FLEISHER, M.D. MARY V. SEEMAN, M.D., F.R.C.P.C. Omaha, Neb. Toronto, Ont., Canada

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1397 BOOK FORUM

MODELS OF PSYCHIATRY The book is a pleasant and interesting read for those with an interest in this area, although it is written about nonpsychiatric primary care. It certainly would be relevant in training medical students and residents, as well as for use on consulta- The Biopsychosocial Approach: Past, Present, Future, tion-liaison services. This volume provides a good back- edited by Richard M. Frankel, Timothy E. Quill, and Susan H. ground but is more of a well-rounded, thoughtful collection McDaniel. Rochester, N.Y., University of Rochester Press, 2003, than an ideal textbook. For didactic purposes, selected chap- 298 pp., $65.00. ters would do well, including Dr. Engel’s two reprinted papers.

It is certainly remarkable that we still need to be reminded References of the most basic, commonsense perspective that informs 1. Engel GL: The need for a new medical model: a challenge for both the art and science of the practice of clinical medicine. biomedicine. Science 1977; 196:129–136 Even Hippocrates understood the importance of approaching 2. Engel GL: The clinical application of the biopsychosocial the patient in his or her social context and treating the whole model. Am J Psychiatry 1980; 137:535–544 person, and early in the last century Adolf Meyer was teaching WILLIAM M. GREENBERG, M.D. a psychosocial approach to psychiatry. Our growing sophisti- Ramsey, N.J. cation as trained physicians in identifying biological underpinnings of diseases should not supplant our understanding of this approach, yet in an atmosphere of more knowledge Mind, Meaning, and Mental Disorder: The Nature of and more paperwork and less time and less continuity of care Causal Explanation in Psychology and Psychiatry, 2nd with patients, it seems we need reorientation to what our full ed., by Derek Bolton and Jonathan Hill. Oxford, U.K., Oxford field of vision and practice should be. University Press, 2004, 387 pp., $52.50 (paper). This volume celebrates the biopsychosocial approach, so aptly named and so inspiringly taught by Dr. George Engel of In the epistemological spirit of trying to make (human) or- Rochester, N.Y. Sadly, Dr. Engel passed away in 1999, while der out of (natural) disorder, this ambitious book attempts to this book was still in preparation. The relevance of his anti- trace the logic-oriented striving of psychology and psychiatry dote to biomedical reductionism has not diminished since its to apply observation, statistics, and logic to the chaotic phe- introduction many decades ago. Dr. Engel teamed with Dr. nomena of the mentally ill mind. For Americans, the use of John Romano, Chair of the Department of Psychiatry at Uni- British spellings (“behaviour”) and grammatical constructions versity of Rochester School of Medicine, to revolutionize the (“the latter became subject to much philosophical stick”) may medical educational culture there, and their seminal influ- be distracting. For believers in biological underpinnings, de- ence in that city continues to this day. Therefore, it is not sur- scriptions such as, “The person appears to have nothing to be prising that 60% of the chapter authors in this volume are af- depressed about, or can think of no reason to be depressed,” filiated with the University of Rochester School of Medicine or may elicit head-shaking outrage. Clinicians seeking research the University of Rochester. or even anecdotal evidence of best practices will be entirely The biopsychosocial approach formally saw the light of day disappointed because these pages offer abstractions only and in print in Dr. Engel’s landmark 1977 Science paper (1) (repro- no immediately applicable hypotheses or data. However, if the duced in this volume), but he is more remembered in psychi- reader is open-hearted and willing to adventure into the over- atry for his 1979 APA Vestermark Award lecture, which was lapping realms of theoretical philosophy and medical nosol- published in the American Journal of Psychiatry in 1980 (2). In ogy, then there may be surprising rewards at the conclusion. the latter, Dr. Engel used the example of a man experiencing a Initial familiarity with the treatises of Plato, Michel Foucault, myocardial infarction and suffering cardiac arrest during his Karl Jaspers, Wittgenstein, Hume, and Thomas Szasz will be medical workup. Seeing this classic reprinted as this book’s helpful, as will a schematic overview of neurobiology. first chapter is a delight. The following chapter by the three The type of analysis in Mind, Meaning, and Mental Disorder editors, introducing the remaining chapters, delivers some may be demonstrated by a series of reasoning statements and unintended irony, as the three authors misremember just- paradigms. For example, if one’s former happy internal state is presented details of that famous vignette. (For the record, the presumed to be linked to a happy external event, and one’s cur- patient endured repeated unsuccessful attempts by several rent unhappy state cannot be associated with any precedent house officers trying to perform an arterial puncture, not start unhappy stimulus, then could both the happiness and unhap- an intravenous line; and the patient said that he consequently piness be random rather than linked to or evoked by biochem- anticipated further painful fumbling attempts and felt out- istry or genetics or contagion? The conclusion of the authors is rage, then self-blame and impotence, not fear of being alone that only brain processes are genuinely causal. Alternatively and abandoned.) stated, 1) intentionality is critical in the regulation and predic- “Clinical Practice and the Biopsychosocial Approach” is an tion of action, 2) intentional processes are causal, 3) inten- impressive chapter seamlessly written by six authors, nicely tional processes that regulate behavior are encoded in the bringing Dr. Engel’s description of biopsychosocial under- brain, and 4) intentionality involves relativity. The process by standing to an up-to-date approach to therapeutic practices. which the authors arrive at their conclusion is through fording Subsequent chapters address other clinical applications, re- the dense and swirling waters of “meaning-as-culture”; the as- search, educational and administrative perspectives, and, fi- sumptions and methods of science; the conceptualization of nally, some historical background and some brief notes on the self, personality, and uniquely individual cognitive com- comparisons with systems theory and current patient-cen- mitment; the distinctions between “understanding” and “ex- tered and relationship-centered approaches to health care. planation”; and the evolutionary connections between psy-

1398 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005 BOOK FORUM choanalytic interpretation and a prescription pad. (The experience. It is a humbling journey, quite like being thrown seemingly benign linguistic caveats of the first paragraph are into a washing machine with bleach and detergent thrown in. not so superficial as might have been suspected.) As awed as I was by Siddhartha and Jiddu Krishnamurti and This volume has nine major divisions, with at least five sub- Nancy Andreasen, Robert Cloninger’s piecing together of all divisions in each. Tracking the lines of thought requires con- the central thoughts of all the positive philosophers, the in- sistent attention; this is not a book to be casually perused in sights of quantum physics, Kurt Godel’s theorem, and Alan bits and pieces over a protracted period of time. The develop- Turing’s insights leading to the development of today’s comment of the fundamental intellectual pursuit here starts with puters makes the maelstrom we call life coherent and hope- the intent of the authors to elucidate the bases of “action,” ful. It saddened me, though, after finishing this book, to think conceptualized as likely related to neural causation. In this how dangerously ill-equipped and unhealthy most of us model, empathy is merely the subjective re-creation of the working in the mental health field are because we are not electrochemical objective existence of the other person. masters of our own consciousness stages. We are expected to Some of the logic is primarily semantic, and other aspects are lead patients who are lost into the promised land of well-be- firmly rooted in known physics and molecular biology. At the ing when most of us are so unwell and so lost. end, anxiety, schizophrenia, and personality disorders are de- I fear that mastery of this book and its concepts would force constructed, and throughout, the examination is conducted three-fourths of the potential workforce of the mental health with true passion and inspiration. Not for everyone. field to drop out from psychiatric residency, social work, psychology, and nursing programs. Those who survived would so LAURA L. POST, M.D., PH.D. Saipan, MP (USA) dramatically overhaul the way they live and approach mental health work that managed care would melt and vanish like the Wicked Witch of the West. If he had his way, Robert Cloninger would create a world MOOD AND AFFECT harking back to the real meaning of the Arabic word for para- dise: pairidaezza, which means “walled-in-garden,” sans snakes, sans institutional religions. It will also mean that Feeling Good: The Science of Well-Being, by C. Robert three-fourths of all psychotropics will be destroyed and John Cloninger, M.D. New York, Oxford University Press, 2004, 374 Lennon’s “Imagine” will supplant all national anthems. The pp., $35.00. problem would be that people who achieve a level of self- transcendent and illuminative level of wisdom eschew power. The French word mise-en-scene means 1) the arrangement TRUCE T. ORDOÑA, M.D. of actors and scenery on a stage for theatrical production, Davenport, Iowa 2) stage setting, or 3) the physical setting of an action. My mother, whose doctorate was on Shakespeare, introduced me Affect Dysregulation and Disorders of the Self, by Allan at age 7 to Act II, Scene VII, Line 139 of Jacques’ monologue in N. Schore, Ph.D. New York, W.W. Norton, 2003, 300 pp., $45.00. As You Like It, which starts with “All the world’s a stage.” That introduction exploded my myopic view of the horizon This book, part of the Norton Series on Interpersonal Neu- forever and set me on my quest for what the real substance is robiology, is a follow-up to Affect Regulation and the Origin of behind and beneath the mise-en-scene of humans. I wanted the Self (1), which focused on regulation theory and its rela- to probe into what lies underneath people’s “stances.” This tion to affect and attachment and their impact on socioemo- led me into an obsessive search to be a polymath and an au- tional development and psychopathology. The current book todidact going beyond the limiting constriction of the scho- attempts to take the topic farther, with sections on Develop- lastic-Aquinian paradigms of Castilian, colonialist Catholi- mental Affective Neuroscience and Developmental Neuro- cism. When I reviewed C. Robert Cloninger’s bibliography, I psychiatry. The author integrates material from social, psy- realized that someone with a similar background could wend chological, and biological realms into a comprehensive through the same peregrinations and come to the same con- model to provide testable psychobiological models of human clusions as a boy born at almost the exact moment of history brain dysfunction. In the process, he presents research from on an island 10,000 miles away. different psychiatric theoretical perspectives, including psy- I offer the following for those who do not believe in seren- chodynamic, behavioral, and sociological frameworks. The dipity: When I set up my practice in 1973, one of my first pa- ideas presented will stimulate clinicians and experimental tients was a young woman who had just attempted suicide. scientists to explore this important topic. Although the book Throughout the hour and a half of interviewing the woman appears intended for those in clinical practice, it may appeal and her family, her father stared at me. When I sent her to the more to theorists and researchers as a synthesis of important ward, I asked the father why he was staring at me so intensely. concepts, theories, and other data. A companion book is Af- He told me that he and his Marine buddy, who had just died, fect Regulation and the Repair of the Self (2). were hiding on my island the day I was born. “I gave you your Section 1, Developmental Affective Neuroscience, provides name because things looked so hopeless because the Japa- an overview for the book (e.g., regulation theory, perspectives nese were kicking our behinds. I named you ‘Truce.’ ” from others on the topic). The concept of self-regulatory sys- Dr. Cloninger has written a book that is profound, well re- tems is introduced and applied to brain development. Indubi- searched, and well thought through. This is not a book for tably, the development of the infant’s brain occurs in the con- people who are authoritarian or who think they know it all. text of a relationship with another self (e.g., the mother). Thus, Reading and absorbing this book is a painful, soul-wrenching a central theme is the experience-dependent relationship be-

Am J Psychiatry 162:7, July 2005 http://ajp.psychiatryonline.org 1399 BOOK FORUM tween infant and mother for the development of self-regula- lytic observations), through years of applied research, to this tory processes. The work of Bowlby and others on attachment proposal of an integrated model. is integrated with neurobiological findings. In particular, the The model argues for two types of depression but is not the function of the right prefrontal cortex, parent-infant commu- classic binary (i.e., endogenous versus neurotic/reactive) one. nication, and attachment disorders are discussed in detail. In essence, Blatt argues for developmentally determined “an- When available, functional magnetic resonance imaging find- aclitic” or dependent depression (where the depressed indi- ings are included. Section 2, Developmental Neuropsychiatry, vidual’s preoccupations involve themes of abandonment and discusses predispositions to psychiatric disorders. Separate loss) and “introjective” depression (typified by punitive and chapters review how brain development, affect regulation, harsh self-criticism), with differential developmental factors and health might be affected by 1) a secure attachment, 2) re- creating the shared vulnerability to depression. As acknowl- lational trauma, and 3) eventual affect dysregulation. edged by Blatt, the model overlaps with views of other theo- We have several suggestions for the next edition. First, we rists, including Bowlby and Beck (the latter used concepts of would like to see case examples to apply the information, in- “sociotropy” and “autonomy”). terweave the presented theories, and “spell out” the concepts The model is multilayered. The two types are held to reflect in a more relevant manner. This change would make it more developmental disturbances in early parent-child relation- readable and more directly applicable clinically. Second, the ships, frame the experiential world of those who develop de- chapters could benefit from a standardized format, with an in- pression, involve a stress-diathesis congruency model linking troduction, objectives, background information, and a sum- contextual stressful events with depression onset, direct mary of key points. This change would also make it more use- treatment approaches, and allow “depression” to be modeled ful as an educational text. Third, in this complex discussion of parsimoniously. The last issue is perhaps the most worrisome the intersection of many fields and particularly neuroscience, plank in Blatt’s argument. Arguing first that depressive sub- it is surprising that few figures or tables are used to help the typing has failed to be achieved by use of symptoms and that reader. Finally, particular emphasis is placed on the relation “depression” is a continuum ranging from “transient dyspho- between structure and function of the right hemisphere dur- ric” responses to “serious distortions of reality” (p. 29), Blatt ing critical stages of development. Unfortunately, supporting then argues for his dichotomous model as providing a “major neurobiological evidence is weak (e.g., regarding the specific- subtyping of depression” that is both reliable and has “dem- ity of abnormal development of the right prefrontal cortex), onstrated validity” (p. 30). Such claims are not supported by and original references are sometimes overinterpreted. This appropriately designed or definitive analyses. Reviewed anal- pitfall, however, surely reflects the paucity of neuroanatomical yses seek to affirm the model rather than refine it by refutabil- information on the normal development of the primate brain. ity studies. This is an interesting, useful integration of data from many Where are the twin studies to consider the genetic and en- fields—all intersecting in the areas of attachment, affect, and vironmental contribution to such personality styles? Why two pathology—that makes it a very unique reference for clini- constructs as against at least four factors of the Big Five cians who want to explore the theoretical underpinnings of model? Is “anaclitic” not a component domain of neuroticism these areas. It is an ambitious work written by a knowledge- and, as measured, weighted to the well-researched but here able author. It is already good, but with the changes noted it essentially ignored facet of rejection sensitivity? If Blatt’s con- would be outstanding in a second edition. tinuum model is valid, how necessary and sufficient are the constructs to distinguish psychotic from melancholic depres- References sion as well as the depressions experienced by psychology 1. Schore AN: Affect Regulation and the Origin of the Self: The students from those of resilient control subjects? In chapter 8 Neurobiology of Emotional Development. New York, WW (“Therapeutic Implications”), how does the model extrapo- Norton, 1994 late to the world of clinical psychiatry (e.g., the perfectionistic 2. Schore AN: Affect Regulation and the Repair of the Self. New York, WW Norton, 2003 individual with a psychotic depression)? Here it would appear that psychotherapy is viewed as necessary and sufficient. DONALD M. HILTY, M.D. There can be little doubt that developmental experiences BLYTHE A. CORBETT, PH.D. and personality styles increase the risk to certain depressive PIERRE LAVENEX, PH.D. Sacramento, Calif. disorders, shape the clinical picture to some degree, and have some differential impact on varying treatment modalities, and that Blatt’s contribution to developing such a matrix has Experiences of Depression: Theoretical, Clinical, and and will continue to be distinctive. However, his exposition in Research Perspectives, by Sidney J. Blatt. Washington, this book posits the model as all-explanatory and has Pro- D.C., American Psychological Association, 2004, 400 pp., crustean overtones. $49.95. Reference Working in the Yale Psychology Department, Sidney Blatt 1. Freud S: Mourning and melancholia (1917 [1915]), in has tenaciously pursued a formulation of depression for more Complete Psychological Works, standard ed, vol 14. London, than 30 years. This book captures his journey and charts the Hogarth Press, 1957, pp 243–258 development of the model from personal sources (early wres- GORDON PARKER, M.D., PH.D., D.SC. tling with Freud’s “Mourning and Melancholia” [1] and ana- Sydney, Australia

Reprints are not available; however, Book Forum reviews can be downloaded at http://ajp.psychiatryonline.org.

1400 http://ajp.psychiatryonline.org Am J Psychiatry 162:7, July 2005